Detectron2 on Custom Dataset | Object Detection Using Detectron2

Learn about Detectron2, an object detection library now implemented in PyTorch. Detectron2 provides support for the latest models and tasks, increased flexibility to aid computer vision research, and improvements in maintainability and scalability to support production use cases.

Detectron2 is Facebooks new vision library that allows us to easily use and create object detection, instance segmentation, keypoint detection and panoptic segmentation models. It has a simple, modular design that makes it easy to rewrite a script for another data-set.

1- Object Detection
2- Image Segmentation
3- Keypoints Detection
4- Panoptic Segmentation

#Detectron #Detectron2 #ObjectDetection #ImageSegmentation #FPN #FeaturePyramidNetwork #MAskRCNN #AI #ComputerVision #Facebook #ArtificialIntelligence #ML #MachineLearning #DeepLearning #PifordTechnologies #AarohiSingla #Akminder

[rajatvardam]

this channel is really underrated. You are doing a noble job and we appreciate your work so much. Keep up.

[vywyv]

Knowing how to code and knowing how to teach others to code...well done. Thank you.

[Harika-]

This video is very useful to me. Please explain how to create our own dataset..

[sohailali]

Very detailed tutorial. Can you please please make video on how can we evaluate this model?

[SwathiVuppala]

It is detecting only ballon not the person .How it will detect baloon and person (this class present in coco dataset)at the same time.

[Syazaniwhd]

Hi thanks for the video! what did you set the working directory as in your Jupyter Notebook

[warrior_]

Mam is there any way we can get to know the best .pth model and calculate val loss during training just like yolov5.

[olas]

Thank you for the explanation, I have one one question. When visualizing the prediction results,
v = Visualizer(im[:, :, ::-1],
metadata=balloon_metadata,
scale=0.5,


the metmetadata=balloon_metadata which was defined above as balloon_metadata =, do we only define this for training dataset or also for validation dataset?

[bahubalipatil]

how can we get approx size of the object? or is there any way to do it? can it be done while labelling the images if yes then how?

[hiteshvazir]

Nice explanation mam pl add it to deep learning play list so it can refer anytime 😊

[JoelPrabhod]

how to run inference on a folder of images locally from a pretrained model and cfg?

[upendrank]

Mam, i have annotation file in csv format. like what you did it in yolov5 custom dataset video. do i need to convert it to json format mam?

[Mr.Esmaily]

Thx for your great tutorials, I've got a question, can i convert h5 to pth file and vice versa ?

[PhilipSantillo]

When using my own .json file in the COCO format, I get a "TypeError: list indices must be integers or slices, not str." In the block of code that checks if the dataset has been loaded properly

[rhymes_touches_myheart]

Hello mam I have one question if I want to use custom dataset having high resolution and which is already present in the Coco for example object "car" so what are initial step I should perform .

[marianaserban]

Hello,
How can we use this model trained on custom dataset for a flask app? I'm a bit confused. For YOLO we exported best.pt file.
How to proceed for detectron 2?

[RaphaelOnyewotu]

Thank you so much for this tutorial, please can you refer me to the VGG annotation tutorial. I have a custom dataset I'd love to annotate. Please can VGG also do a batch annotation or I have to do it manually for each Image. I have about 9000 images

[aboudezoa]

video title for object detection and code for instance segmentation

[jakubkahoun]

Maybe its the problem, you have only positive dataset and no negative one?

[PhilipSantillo]

Hello, I'm getting the error "ModuleNotFoundError: No module named 'detectron2.engine' ", I believe it's due to the cloning error that I also received "fatal: destination path 'detectron2' already exists and is not an empty directory." I'm running Jupiter notebook in Anaconda environment, and before running I, conda -c condo-forge detectron2 the library. any installation advice?