Computer Vision Tutorial: Implementing Mask R-CNN for Image Segmentation

Pulkit Sharma 25 Jun, 2024
14 min read


I am fascinated by self-driving cars. The sheer complexity and mix of different computer vision techniques that go into building a self-driving car system is a dream for a data scientist like me.

So, I set about trying to understand the computer vision technique behind how a self-driving car potentially detects objects. A simple object detection framework might not work because it simply detects an object and draws a fixed shape around it.

That’s a risky proposition in a real-world scenario. Imagine if there’s a sharp turn in the road ahead and our system draws a rectangular box around the road. The car might not be able to understand whether to turn or go straight. That’s a potential disaster!

Mask R CNN

Instead, we need a technique that can detect the exact shape of the road so our self-driving car system can safely navigate the sharp turns as well.


  • Mask R-CNN is a state-of-the-art framework for Image Segmentation tasks
  • We will learn how Mask R-CNN works in a step-by-step manner
  • We will also look at how to implement Mask RCNN in Python and use it for our own images

The latest state-of-the-art framework that we can use to build such a system? That’s Mask R-CNN!

So, in this article, we will first quickly look at what image segmentation is. Then we’ll look at the core of this article – the Mask R CNN framework. Finally, we will dive into implementing our own Mask R CNN model in Python. with that also you will clarified about the Mask R CNN Pytorch Implementation So,Let’s begin!

What is Mask R-CNN?

Mask R-CNN, which stands for Mask Region-based Convolutional Neural Network, is a deep learning model that tackles computer vision tasks like object detection and instance segmentation. It builds upon an existing architecture called Faster R-CNN.

Two main types of image segmentation of Mask R-CNN

Semantic Segmentation involves categorizing every pixel in an image into a distinct class. Picture a scene where individuals are strolling along a road. Semantic segmentation is able to categorize all the pixels belonging to individuals as a single group, despite there being numerous people.

Instance Segmentation: This method goes beyond semantic segmentation by not only classifying pixels but also differentiating between objects of the same class. In the same image with people walking, instance segmentation would create a separate mask for each individual person.

Also Read: A Comprehensive Tutorial to learn Convolutional Neural Networks

A Brief Overview of Image Segmentation

We learned the concept of image segmentation in part 1 of this series in a lot of detail. We discussed what is image segmentation and its different techniques, like region-based segmentation, edge detection segmentation, and segmentation based on clustering.

I would recommend checking out that article first if you need a quick refresher (or want to learn image segmentation from scratch).

I’ll quickly recap that article here. Image segmentation creates a pixel-wise mask for each object in the image. This technique gives us a far more granular understanding of the object(s) in the image. The image shown below will help you to understand what image segmentation is:

blood cells

Here, you can see that each object (which are the cells in this particular image) has been segmented. This is how image segmentation works.

We also discussed the two types of image segmentation: Semantic Segmentation and Instance Segmentation. Again, let’s take an example to understand both of these types:

types of image segmentation

All 5 objects in the left image are people. Hence, semantic segmentation will classify all the people as a single instance. Now, the image on the right also has 5 objects (all of them are people). But here, different objects of the same class have been assigned as different instances. This is an example of instance segmentation.

Part one covered different techniques and their implementation in Python to solve such image segmentation problems. In this article, we will be implementing a state-of-the-art image segmentation technique called Mask RCNN to solve an instance segmentation problem.

Also Read: A Step-by-Step Guide to Image Segmentation Techniques

Understanding Mask R-CNN

Mask R CNN is basically an extension of Faster R-CNN. Faster R-CNN is widely used for object detection tasks. For a given image, it returns the class label and bounding box coordinates for each object in the image. So, let’s say you pass the following image:

cat and dog

The Fast R-CNN model will return something like this:

cat and dog: object detection

The Mask R-CNN framework is built on top of Faster R-CNN. So, for a given image, Mask R-CNN, in addition to the class label and bounding box coordinates for each object, will also return the object mask.

Let’s first quickly understand how Faster R-CNN works. This will help us grasp the intuition behind Mask R-CNN as well.

  • Faster R-CNN first uses a ConvNet to extract feature maps from the images
  • These feature maps are then passed through a Region Proposal Network (RPN) which returns the candidate bounding boxes
  • We then apply an RoI pooling layer on these candidate bounding boxes to bring all the candidates to the same size
  • And finally, the proposals are passed to a fully connected layer to classify and output the bounding boxes for objects

Once you understand how Faster R-CNN works, understanding Mask R-CNN will be very easy. So, let’s understand it step-by-step starting from the input to predicting the class label, bounding box, and object mask.

Backbone Model

Similar to the ConvNet that we use in Faster R-CNN to extract feature maps from the image, we use the ResNet 101 architecture to extract features from the images in Mask RCNN. So, the first step is to take an image and extract features using the ResNet 101 architecture. These features act as an input for the next layer.

Region Proposal Network (RPN)

Now, we take the feature maps obtained in the previous step and apply a region proposal network (RPM). This basically predicts if an object is present in that region (or not). In this step, we get those regions or feature maps which the model predicts contain some object.

Region of Interest (RoI)

The regions obtained from the RPN might be of different shapes, right? Hence, we apply a pooling layer and convert all the regions to the same shape. Next, these regions are passed through a fully connected network so that the class label and bounding boxes are predicted.

Till this point, the steps are almost similar to how Faster R-CNN works. Now comes the difference between the two frameworks. In addition to this, Mask R-CNN also generates the segmentation mask.

For that, we first compute the region of interest so that the computation time can be reduced. For all the predicted regions, we compute the Intersection over Union (IoU) with the ground truth boxes. We can computer IoU like this:

IoU = Area of the intersection / Area of the union

Now, only if the IoU is greater than or equal to 0.5, we consider that as a region of interest. Otherwise, we neglect that particular region. We do this for all the regions and then select only a set of regions for which the IoU is greater than 0.5.

Let’s understand it using an example. Consider this image:


Here, the red box is the ground truth box for this image. Now, let’s say we got 4 regions from the RPN as shown below:


Here, the IoU of Box 1 and Box 2 is possibly less than 0.5, whereas the IoU of Box 3 and Box 4 is approximately greater than 0.5. Hence. we can say that Box 3 and Box 4 are the region of interest for this particular image whereas Box 1 and Box 2 will be neglected.

Next, let’s see the final step of Mask RCNN.

Segmentation Mask

Once we have the RoIs based on the IoU values, we can add a mask branch to the existing architecture. This returns the segmentation mask for each region that contains an object. It returns a mask of size 28 X 28 for each region which is then scaled up for inference.

Again, let’s understand this visually. Consider the following image:

sample image for segmentation

The segmentation mask for this image would look something like this:

masks using Mask R-CNN

Here, our model has segmented all the objects in the image. This is the final step in Mask R-CNN where we predict the masks for all the objects in the image.

Keep in mind that the training time for Mask R-CNN is quite high. It took me somewhere around 1 to 2 days to train the Mask R-CNN on the famous COCO dataset. So, for the scope of this article, we will not be training our own Mask R CNN model.

We will instead use the pretrained weights of the Mask R-CNN model trained on the COCO dataset. Now, before we dive into the Python code, let’s look at the steps to use the Mask R CNN model to perform instance segmentation.

Steps to implement Mask R-CNN

It’s time to perform some image segmentation tasks! We will be using the mask rcnn framework created by the Data scientists and researchers at Facebook AI Research (FAIR).

Let’s have a look at the steps which we will follow to perform image segmentation using Mask RCNN.

Step 1: Clone the repository

First, we will clone the mask rcnn repository which has the architecture for Mask R-CNN. Use the following command to clone the repository:

git clone

Once this is done, we need to install the dependencies required by Mask RCNN.

Step 2: Install the dependencies

Here is a list of all the dependencies for Mask R CNN:

  • numpy
  • scipy
  • Pillow
  • cython
  • matplotlib
  • scikit-image
  • tensorflow>=1.3.0
  • keras>=2.0.8
  • opencv-python
  • h5py
  • imgaug
  • IPython

You must install all these dependencies before using the Mask R-CNN framework.

Step 3: Download the pre-trained weights (trained on MS COCO)

Next, we need to download the pretrained weights. You can use this link to download the pre-trained weights. These weights are obtained from a model that was trained on the MS COCO dataset. Once you have downloaded the weights, paste this file in the samples folder of the Mask_RCNN repository that we cloned in step 1.

Step 4: Predicting for our image

Finally, we will use the Mask R-CNN architecture and the pretrained weights to generate predictions for our own images.

Once you’re done with these four steps, it’s time to jump into your Jupyter Notebook! We will implement all these things in Python and then generate the masks along with the classes and bounding boxes for objects in our images.

What is Convolutional Neural Networks (CNN)?

Convolutional Neural Networks (CNNs) are a type of deep learning neural network that excel at analyzing visual imagery like photos and videos. They are inspired by the structure of the animal visual cortex, which is the part of the brain responsible for processing vision.

Implementing Mask R-CNN in Python

Sp, are you ready to dive into Python and code your own image segmentation model? Let’s begin!

To execute all the code blocks which I will be covering in this section, create a new Python notebook inside the “samples” folder of the cloned Mask_RCNN repository.

Let’s start by importing the required libraries:

import os
import sys
import random
import math
import numpy as np
import matplotlib
import matplotlib.pyplot as plt

# Root directory of the project
ROOT_DIR = os.path.abspath("../")

import warnings

# Import Mask RCNN
sys.path.append(ROOT_DIR)  # To find local version of the library
from mrcnn import utils
import mrcnn.model as modellib
from mrcnn import visualize
# Import COCO config
sys.path.append(os.path.join(ROOT_DIR, "samples/coco/"))  # To find local version
import coco

%matplotlib inline

Next, we will define the path for the pretrained weights and the images on which we would like to perform segmentation:

# Directory to save logs and trained model
MODEL_DIR = os.path.join(ROOT_DIR, "logs")

# Local path to trained weights file
COCO_MODEL_PATH = os.path.join('', "mask_rcnn_coco.h5")

# Download COCO trained weights from Releases if needed
if not os.path.exists(COCO_MODEL_PATH):

# Directory of images to run detection on
IMAGE_DIR = os.path.join(ROOT_DIR, "images")

If you have not placed the weights in the samples folder, this will again download the weights. Now we will create an inference class which will be used to infer the Mask R-CNN model:

class InferenceConfig(coco.CocoConfig):
    # Set batch size to 1 since we'll be running inference on
    # one image at a time. Batch size = GPU_COUNT * IMAGES_PER_GPU
    GPU_COUNT = 1

config = InferenceConfig()
view hosted with ❤ by GitHub
mask rcnn model summary

What can you infer from the above summary? We can see the multiple specifications of the Mask R-CNN model that we will be using.

So, the backbone is resnet101 as we have discussed earlier as well. The mask shape that will be returned by the model is 28X28, as it is trained on the COCO dataset. And we have a total of 81 classes (including the background).

We can also see various other statistics as well, like:

  • The input shape
  • Number of GPUs to be used
  • Validation steps, among other things.

You should spend a few moments and understand these specifications. If you have any doubts regarding these specifications, feel free to ask me in the comments section below.

Loading Weights

Next, we will create our model and load the pretrained weights which we downloaded earlier. Make sure that the pretrained weights are in the same folder as that of the notebook otherwise you have to give the location of the weights file:

# Create model object in inference mode.
model = modellib.MaskRCNN(mode="inference", model_dir='mask_rcnn_coco.hy', config=config)

# Load weights trained on MS-COCO
model.load_weights('mask_rcnn_coco.h5', by_name=True)

Now, we will define the classes of the COCO dataset which will help us in the prediction phase:

# COCO Class names
class_names = ['BG', 'person', 'bicycle', 'car', 'motorcycle', 'airplane',
               'bus', 'train', 'truck', 'boat', 'traffic light',
               'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird',
               'cat', 'dog', 'horse', 'sheep', 'cow', 'elephant', 'bear',
               'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie',
               'suitcase', 'frisbee', 'skis', 'snowboard', 'sports ball',
               'kite', 'baseball bat', 'baseball glove', 'skateboard',
               'surfboard', 'tennis racket', 'bottle', 'wine glass', 'cup',
               'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple',
               'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza',
               'donut', 'cake', 'chair', 'couch', 'potted plant', 'bed',
               'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote',
               'keyboard', 'cell phone', 'microwave', 'oven', 'toaster',
               'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors',
               'teddy bear', 'hair drier', 'toothbrush']

Let’s load an image and try to see how the model performs. You can use any of your images to test the model.

# Load a random image from the images folder
image ='sample.jpg')

# original image
sample image for image segmentation

This is the image we will work with. You can clearly identify that there are a couple of cars (one in the front and one in the back) along with a bicycle.

Making Predictions

It’s prediction time! We will use the Mask R-CNN model along with the pretrained weights and see how well it segments the objects in the image. We will first take the predictions from the model and then plot the results to visualize them:

# Run detection
results = model.detect([image], verbose=1)

# Visualize results
r = results[0]
visualize.display_instances(image, r['rois'], r['masks'], r['class_ids'], class_names, r['scores'])
mask rcnn result

Interesting. The model has done pretty well to segment both the cars as well as the bicycle in the image. We can look at each mask or the segmented objects separately as well. Let’s see how we can do that.

I will first take all the masks predicted by our model and store them in the mask variable. Now, these masks are in the boolean form (True and False) and hence we need to convert them to numbers (1 and 0). Let’s do that first:

mask = r['masks']
mask = mask.astype(int)



This will give us an array of 0s and 1s, where 0 means that there is no object at that particular pixel and 1 means that there is an object at that pixel. Note that the shape of the mask is similar to that of the original image (you can verify that by printing the shape of the original image).

However, the 3 here in the shape of the mask does not represent the channels. Instead, it represents the number of objects segmented by our model. Since the model has identified 3 objects in the above sample image, the shape of the mask is (480, 640, 3). Had there been 5 objects, this shape would have been (480, 640, 5).

We now have the original image and the array of masks. To print or get each segment from the image, we will create a for loop and multiply each mask with the original image to get each segment:

for i in range(mask.shape[2]):
    temp ='sample.jpg')
    for j in range(temp.shape[2]):
        temp[:,:,j] = temp[:,:,j] * mask[:,:,i]
segment result 1
segment result 2
segment result 3

This is how we can plot each mask or object from the image. This can have a lot of interesting as well as useful use cases. Getting the segments from the entire image can reduce the computation cost as we do not have to preprocess the entire image now, but only the segments.


Below are a few more results which I got using our Mask R-CNN model:

mask rcnn result 1
mask rcnn result 2
mask rcnn result 3
mask rcnn result 4

Looks awesome! You have just built your own image segmentation model using Mask R-CNN – well done.

How to implement Mask R-CNN using PyTorch:

Install Required Libraries:

  • Ensure you have torch, torchvision, and pycocotools installed. If not, install them using:
pip install torch torchvision pycocotools
  1. Load the Pre-trained Mask R-CNN Model:
    • PyTorch’s torchvision library provides a pre-trained Mask R-CNN model.
  2. Prepare the Dataset:
    • Mask R-CNN is typically trained on the COCO dataset, but you can prepare your own dataset following the COCO format.
  3. Inference with Mask R-CNN:
    • Load an image, preprocess it, and pass it through the model to get the predictions.

Here’s a sample code snippet demonstrating these steps:

import torch
import torchvision
from PIL import Image
import matplotlib.pyplot as plt
import torchvision.transforms as T

# Load a pre-trained Mask R-CNN model
model = torchvision.models.detection.maskrcnn_resnet50_fpn(pretrained=True)
model.eval()  # Set the model to evaluation mode

# Load an image
img_path = 'path_to_your_image.jpg'
img ="RGB")

# Preprocess the image
transform = T.Compose([T.ToTensor()])
img = transform(img)

# Add a batch dimension
img = img.unsqueeze(0)

# Perform inference
with torch.no_grad():
    prediction = model(img)

# Visualize the results
def plot_image(image, masks, boxes):
    fig, ax = plt.subplots(1, figsize=(12,9))
    for i in range(len(masks)):
        mask = masks[i].cpu().numpy()
        box = boxes[i].cpu().numpy()
        ax.imshow(mask, alpha=0.5)
        rect = plt.Rectangle((box[0], box[1]), box[2] - box[0], box[3] - box[1], fill=False, color='red')

# Convert image back to numpy
img_np = img.squeeze().permute(1, 2, 0).cpu().numpy()

# Plot the image with masks and bounding boxes
plot_image(img_np, prediction[0]['masks'], prediction[0]['boxes'])


Exploring the table of contents provides a comprehensive understanding of image segmentation, particularly focusing on the implementation of Mask R-CNN in Python. The step-by-step guide outlined elucidates the process from cloning the repository to predicting for customized images. Through understanding Mask R-CNN and following the prescribed steps, users can effectively utilize pre-trained weights and execute segmentation tasks efficiently. This structured approach empowers practitioners to harness the capabilities of advanced segmentation techniques, enhancing their proficiency in image analysis and fostering innovation in various fields reliant on computer vision technologies. Also, you will how to implement mask r CNN pytorch.

Key Takeaways :

  • Mask R-CNN extends Faster R-CNN to perform instance segmentation, providing object detection, classification, and pixel-wise masks.
  • The framework uses a backbone model (ResNet 101) for feature extraction, followed by Region Proposal Network and Region of Interest alignment.
  • Implementation involves using pre-trained weights, setting up the environment, and running inference on images using Python libraries.
  • Mask R-CNN can be implemented using both TensorFlow/Keras and PyTorch, with pre-trained models available in both frameworks.

Frequently Asked Questions

Q1. What is mask R-CNN used for?

A. Mask R-CNN is used for object detection, segmentation, and instance segmentation in images, identifying objects and delineating their exact shapes.

Q2. What is the difference between mask R-CNN and CNN?

A. CNN (Convolutional Neural Network) is primarily used for image classification and feature extraction, while Mask R-CNN extends Faster R-CNN by adding a branch for predicting segmentation masks on each detected object.

Q3. What does the R-CNN stand for?

A. R-CNN stands for Region-based Convolutional Neural Network, which focuses on identifying regions of interest in an image for object detection.

Q4. What are mask R-CNN loss functions?

A. Mask R-CNN uses multiple loss functions: classification loss for object categories, bounding box regression loss for object localization, and mask loss for pixel-wise object segmentation.

Pulkit Sharma 25 Jun, 2024

My research interests lies in the field of Machine Learning and Deep Learning. Possess an enthusiasm for learning new skills and technologies.

Frequently Asked Questions

Lorem ipsum dolor sit amet, consectetur adipiscing elit,

Responses From Readers


Vaibhav Mathur
Vaibhav Mathur 22 Jul, 2019

Hi Pulkit, great article I was looking for the same. Thanks for sharing.

Praketa Saxena
Praketa Saxena 22 Jul, 2019

hey really nice article...I am trying to recognize and classify SSD boards is this something i can employ in my research

Pankaj Jainani
Pankaj Jainani 23 Jul, 2019

Hello Pulkit, Yet again, nice way to explain the key concepts.

sruthin 23 Jul, 2019

can you explain in the same way for yolo and single shot multi box detector.

Rohan 25 Jul, 2019

I'm training MaskRCNN for custom dataset using pretrained yolo weights, and default configuration with changes in class names, I have trained network heads for 90epoch and the entire network later for (125-90)epochs with 100 iterations per epoch, the loss is jumping and the model is not converging. What do I do

Dan 29 Jul, 2019

This is awesome content Pulkit. Thank you so much for your contribution to enable people like me try out this code. Your detailed step by step explanations are very helpful. I am very grateful. Please continue with this. I really appreciate it! Just one question. When I do the printing loop I get 3 images for each object. First a green one then a blue one and finally the mask. Have I done something wrong? Best regards

GEET 31 Jul, 2019

How to test the unseen dataseen dataset and evaluate mAP on mrcnn

Sunita Roy
Sunita Roy 04 Aug, 2019

Hi Pulkit, I want to create a model for face semantic segmentation. Can you please explain the steps in short.

Prisilla 08 Aug, 2019

Hi Pulkit, Does Mask R-CNN framework not work accurately for finding the inference of aerial images? It was not able to detect the bird instead assumed as kite and in other case for a group of building it gave a result as train. I used the pre-trained labels. Can you please let me know how to label the objects of any given image.

Sruti Bhattacharjee
Sruti Bhattacharjee 13 Aug, 2019

Hi Pulkit, Nice article. I have some confusion regarding saving the data. I want to use my own dataset where I have images and their annotation JSON files. Can you elaborate about it?

Arnold 16 Aug, 2019

Hi there Pulkit, This is a great knowledge, do you have a next step by step tutorial on how to train your own datasets in Mask R-CNN. It will be a great context for us who what to learn a lot .

Hansy 24 Aug, 2019

Hi Pulkit Thanks very much for useful Tutorial. I am doing my research on Lungs nodules detection from CT scan. Can I use this Mask R CNN for nodules segmentation provided that I have ground truth ( annotated images) from radiologist?

Farhad 26 Aug, 2019

Pulkit Sharma, thanks for your awesome article. Can I use this framework for skin disease ulcers segmentation and detection? Thanks,

Afshin 01 Sep, 2019

Thank you for your great tutorial. I have a question about runtime. If we want to report the time takes to get result, do we report the time for training our train set or getting the result from test set? Do these two separated or not? Could you please explain about the time concept in the implementation such a deep model?

Kiprono Elijah KOECH
Kiprono Elijah KOECH 02 Sep, 2019

I always like your blogs. Thank you for this presentation. I have tried to train Mask R-CNN and I have failed until I started reading your articles. I wish to know how to develop Intersection over Union (IoU) metric given ground truth and prediction contours.

oacemi 03 Sep, 2019

Hi! How can i save the each masked image as png or jpg after writing this code: "for i in range(mask.shape[2]): temp ='sample.jpg') for j in range(temp.shape[2]): temp[:,:,j] = temp[:,:,j] * mask[:,:,i] plt.figure(figsize=(8,8)) plt.imshow(temp)" I want to save all segmented masks from image without black background. Just segmented masks(images)

Kiprono Elijah KOECH
Kiprono Elijah KOECH 07 Sep, 2019

Very good article. Have you written on mask R-CNN on own dataset?

Akmaral 12 Sep, 2019

Hi, Thank you for great explanation which is very helpful. I am wonder whether Mask R-CNN performs well with bin-picking challenge (segmentation of multiple objects of the same type)? I would appreciate your help and guidance. Thank you.

Akmaral 12 Sep, 2019

Hi, Thank you for this great explanation. I am wondering whether Mask R-CNN performs well in case of bin-picking challenge (segmentation of multiple objects of the same type + occluded scene). I would appreciate your guidance. Thank you.

Hugo Santos
Hugo Santos 19 Sep, 2019

Hi! Have you ever tryed to use an aerial image dataset (from an UAV) containing trees (or vegetation in general) to detect and segment them? I've tryed to do transfer learning and the results are not so good. I have almost 200 images to use as training/test dataset. I know that it is few images, but what you could recommend to acquire best results? Thanks in advance!

Abhishek Kumar Prajapati
Abhishek Kumar Prajapati 25 Sep, 2019

I've already trained FRCNN model on a dataset. Can I anyhow use the trained weights (i.e. for FRCNN) in mask-RCNN for segmentation? Thanks in advance.

Abi 26 Sep, 2019

Hi, What is the criteria for setting the number of steps in each iterations?

Yass 06 Oct, 2019

Hi Pulkit This is really good article....thank you I have research paper for final year . so can you help me please? I want to use dataset of blood (for example ) but I don't know how to run it in mask R-CNN. I follow article that explains steps to do that. but I can't download and convert it from xml to csv. Thanks in advance

Shivam 06 Oct, 2019

Hi Pulkit, I wanted to know if there's any way to just mask out the elements without comparing them with labels? Basically I mean, a decent way of segmenting the objects out of the image to identify that there is some solid object in the image. I am not concerned about the labels.

Andrés Felipe Castaño Morales
Andrés Felipe Castaño Morales 07 Oct, 2019

Man, you are the best. I'm so grateful man, thanks!

Rabiu 09 Oct, 2019

Thank you for the wonderful guide. I am fresh starter in the area, I am using jupyter on windows 10, my question is how can I download and install the dependencies mentioned in step 2. Step 2: Install the dependencies Here is a list of all the dependencies for Mask R-CNN: numpy scipy Pillow cython matplotlib scikit-image tensorflow>=1.3.0 keras>=2.0.8 opencv-python h5py imgaug IPython

Imran Ahmed
Imran Ahmed 14 Oct, 2019

What should I do if I want to make a prediction on satellite images? I replaced the image "sample.jpg" with my own satellite image, but it isn't working, saying ***No instance to display" please give a guideline to use this for segmenting satellite imagery.

Pawan 17 Oct, 2019

i would try to classify the standing person and sitting person by using object detection technique but i fail to classify this. so ,what technique we will use on it ?

Durga 18 Oct, 2019

Thank you for making the concept very clear. Could you also come up with a tutorial on creating custom dataset masks and using them for instance segmentation.

thanh 24 Oct, 2019

I had an error below: ValueError: Layer #389 (named "mrcnn_bbox_fc"), weight has shape (1024, 324), but the saved weight has shape (1024, 8). Please help me..

achim 28 Oct, 2019

Pulkit, this is an interesting tutorial about image segmentation. Would you be able to point us to another tutorial (or to write one yourself) that shows how to generate test and training datasets for image segmentation? What are the best tools to generate the masks needed? How are they applied? Is it feasible to do this or do you always need an army of annotators? Thanks again for your work, Achim

Pantelis 28 Oct, 2019

Hi PULKIT SHARMA, thank you for your effort. I would like to ask you if the repository that you provided has changed since you wrote the article. After following your instructions, I can't run the script. I get the following error: Using TensorFlow backend. Traceback (most recent call last): File "C:/Users/kaniourasp/Downloads/Thesis/mask-rcnn/samples/", line 26, in import coco File "C:\Users\kaniourasp\Downloads\Thesis\mask-rcnn\samples/coco\", line 42, in from pycocotools.coco import COCO ModuleNotFoundError: No module named 'pycocotools' After browsing through my folder, I couldn't find the pycocotools module.

Bhaskar 28 Oct, 2019

Hello Sir, I want to count and display the number of same objects. please provide code for this

Monika 31 Oct, 2019

hello sir, thank you for the blog, it is very helpful as I am new to deep learning field. But I have been struggling to plot its training vs validation loss and accuracy plot. can you help to solve the problem?

Honey Bansal
Honey Bansal 05 Nov, 2019

I am training a model for detection purpose , I have 2 lakhs of images in my dataset, so how can we decide the timesteps and number of epochs for Mask RCNN to train. Thanks in advance!!! Also, how to use GPU for MASK RCNN training, If I set GPU_count=0 then what will happen? or if I set GPU_COUNT=1 then what will happen?

Jennifer 09 Nov, 2019

How can i segment a histopathological image, i only have its ground truth annotations in json format

Lavanya 19 Nov, 2019

how to customize the code for single class like person

Thomas Kaar
Thomas Kaar 20 Nov, 2019

Great Article!

Thomas Mai
Thomas Mai 23 Nov, 2019

Thanks for the great guide. I am new to this area, can you tell me how can I obtain the coordinates of the segmented object? Thanks in advance.

Vikram Natraj
Vikram Natraj 28 Nov, 2019

Hi. Great Article. Could you tell me if this can be used to detect fake images/tampering in images? I want to build a classifier using your model to find out whether an image is take or not. Can it be used ? Please let me know.

Mohammadreza 02 Dec, 2019

Hi dear Pulkit! I'm work on deep learning project that has two part. first part is object detection that use Faster RCNN pretrained model has .CKPT format second part use the detected object and bounding boxes. your article is awesome and I want to use Mask RCNN instead of faster rcnn. but all the time I work with the second part and I don't know how to use Mask Rcnn after download it.

UDDIN 04 Dec, 2019

Hello, Pulkit Sharma. Thanks for your great article. Can i use this to detect Computer small parts like RAM, processor, motherboard etc. ? I want to classify every parts separately and count them. Please let me know. Thanks in advance.

mani 05 Dec, 2019

Hello pulkit sharma, Thanks for your article. But I'm doing project on "artichoke flower image" can i use this to detect the artichoke flower. please let me know.. Thank you ..

Shreyas Talole
Shreyas Talole 05 Dec, 2019

These steps do not work on windows as default c compiler is not present in windows. Help to solve this problem at earliest

Kadir 06 Dec, 2019

Thank you so much. This is a great article and very easy explanation. I got a "ModuleNotFoundError: No module named 'pycocotools' " and solved. If anybody struggling with the same problem on Windows, conda, I am sharing the solution as below. Run the following codes in order conda install git pip install "git+"

Mali 07 Dec, 2019

Thanks for this tutorial this hepls me a lot. I need to ask one question. How can i make the masking white. I want objects background white. Thanks for helping.

Sergio Gregori
Sergio Gregori 11 Dec, 2019

hi Pulkit Good job! congrats! I have a question, I need to create a real-time weed detector. Can I use this concept to make my project? In other hand do you know a course where i can learn how to train a model from scratch for mask-rcnn? Regards

Pandito HARS
Pandito HARS 13 Dec, 2019

Thanks for your tutorial sir its very helpful. Is there any chance if i use the MS-COCO pre trained model, and then for the testing we only visualize vehicles classes (motorcycle,car,bus,truck)? Maybe u have any suggestion to the code changing, thanks before.

divya singh
divya singh 18 Dec, 2019

hi pulkit thanx for this article.. have you any idea how can claculate confusion metrix for Mask RCNN.... please help me...

Yusra Khalid
Yusra Khalid 23 Dec, 2019

hello pulkit, This is such a Great article esp for beginners. My question is if i have to apply this for emotion recognition of instructors in classroom environment, how would i configure it. I have 200-300 images in total of 5 classes.

hr 312
hr 312 23 Dec, 2019

Hey! Great Demonstration. Just one question, What's the inference time for MaskRCNN and what according to you is faster in terms of inference and performs good on document data? Size of the dataset is 1000 training samples.

shankar 25 Dec, 2019

Dear Pulkit, well done mate !! Mask-RCNN first proposes regions of interest. Then for each RoI, it generates its feature vector. Then it classifies these feature vectors into C classes, and also gives the corresponding bounding boxes+masks. For my research, in addition to the box + mask + class_label, I need the feature vector of each of the detected objects. Can you guide me (via this forum or my email) to extract the feature vectors of objects? I shall be grateful :)

Gurdev Singh
Gurdev Singh 06 Jan, 2020

Very nice article, but I have a question. How can we get the bounding box for each mask? I am in need of it.

Jondal 07 Jan, 2020

Awesome Job Pulkit Sharma! I am still quite new to the subject. Currently I am working on a dataset with different types of fish. I have labeled images for every fish type, but without bounding boxes for the training data like the PASCAL or COCO data set. Sometimes the fish species I am looking for are a bit small in the pictures, but the training data set is large enough to compensate for this. Can I still apply the model to my data so that a fish is recognized and found?

dnrg 14 Jan, 2020

"Step 3: Download the pre-trained weights (trained on MS COCO) Next, we need to download the pretrained weights. You can use this link to download the pre-trained weights. " There are 4 resources under that link. Which resource is the weights? Wish ML authors would be less sloppy

Cesar Navarro
Cesar Navarro 22 Jan, 2020

Thank you for your article! I've followed it but I encountered problems with version 2 of tf I have installed on my laptop. I did some changes and now it is working like a treat. I've made the following changes in Mask_RCNN\mrcnn\ to accommodate tf v2: ---> line 341 return tf.log(x) / tf.log(2.0) replaced by return tf.math.log(x) / tf.math.log(2.0) ---> lines 720 y 758 keep = tf.sets.set_intersection(tf.expand_dims(keep, 0), replaced by keep = tf.sets.intersection(tf.expand_dims(keep, 0), ---> lines 722 y 760 keep = tf.sparse_tensor_to_dense(keep)[0] replaced by keep = tf.sparse.to_dense(keep)[0]

Cesar Navarro
Cesar Navarro 22 Jan, 2020

I've forgotten in my previous post when using with tf v2: Additionally to the mentioned updates I needed to replace in line 772: tf.to_float(tf.gather(class_ids, keep))[..., tf.newaxis], by tf.dtypes.cast(tf.gather(class_ids, keep), 'float32', name='Cast_Float')[..., tf.newaxis], 'float64' did not work for me but 'float32' did :)

Natta 27 Jan, 2020

Where is the White-blue striped car and bicycle download link? Or you want us to use what ever image to see the result?

Alexey 29 Jan, 2020

Thank you very much for the article! With the recognition of objects, all is well mask built and displayed. Have a question, can you tell me? How to check the presence of an object in a pre-drawn rectangle in the picture.

Ilkin 18 Feb, 2020

Hi Pulkit, Thanks for nice tutorials. I read both of the segmentation articles. I'm wondering if segmentation is better than detection in terms of accuracy. Also, I recently encountered a segmentation algorithm called DeepLab. It would be nice if you can compare it with Mask RCNN as next step

Mathew Tho
Mathew Tho 21 Feb, 2020

Thank you Pulkit for this very well explained tutorial! I wondered if you already have writen a tuto explaining how to make our own dataset (COCO format, but with our own photos and classes)? And accessorely how to re-train the RCNN to take into account this new dataset. I know that the training is pretty time-consuming, but that's life... Anyway, thank you for you vulgarisation job ;-) Mathieu from France

Hamed Salem Mounchili Salim
Hamed Salem Mounchili Salim 05 Mar, 2020

Hi Pulkit! Thanks for explaining how the mask R-CNN works. I need some advice on object detection in Python. I am trying to recognize an object in a live image. I have a camera that makes a live video of an individual's mouth. I would like to mark the contours of the tooth with lines and would like that when the camera moves or the mouth, these contours (marked lines) also move to mark a tooth always. My question would be how could I program something like this in Python. I'm really blue when it comes to image processing and object detection, which is why I turn to you for help. I have read a lot of tutorial so far but unfortunately I still can't do it. Help me please! Do you think that it can works with the mask R-CNN ?

Rod 08 Mar, 2020

Hi Pulkit, Do you have a consulting business? I have a project I would like for you to consider. Please email me directly. [email protected]

Ashtin 12 Mar, 2020

Can you tell me the number of images you used for training, validation and testing? Or, how many images are required for training in general?

Tejas 12 Mar, 2020

Hi. I want to know, what is the sample size you used for training? Also what is the minimum size required in general? As I have a data-set with 10 classes. I am using Mask-RCNN. Just want to know number of samples that are required for training.

Pavan K R
Pavan K R 15 Mar, 2020

This was extremely helpful. I think no one has ever done such a detailed step by step explanation.

Tharun. 06 Dec, 2021

Hi pulkit. I am working on an object detection task in microscopic videos and images. I have manually generated Segmentation images by using 3D slicer. So, I'm using segmented masks as ground truth data. So, how can I approach with Mask R-CNN to detetct the objects in my use case?

Tharun 06 Dec, 2021

Hi pulkit. I am working on an object detection task in microscopic videos and images. I have manually generated Segmentation images by using 3D slicer. So, I'm using segmented masks as ground truth data. So, how can I approach with Mask R-CNN to detetct the objects in my use case?

Francisco 06 Apr, 2022

Hello!! I am implementing the maskrcnn algorithm to segment the detected object but when executing model.load_weights(COCO_MODEL_PATH, by_name=True) it returns the following error: ValueError: Layer #389 (named "mrcnn_bbox_fc"), weight has shape (1024, 4), but the saved weight has shape ( 1024, 324). And I don't know how to solve it, I would appreciate your help. From already thank you very much!!

Jhe Gwapa
Jhe Gwapa 09 Apr, 2022

Thank you for this very helpful article. But at cropping the segmented object, it outputs separate image per mask. Can I make only one image for all the detected or cropped object? How? Please help me

bouh 11 Dec, 2023

hello thank you for explantation please if i need to train MRCNN for my dataset what i need to do that not classes in the MS coco ?