Add Functionality for Object Detection

harsh2ai / Trinetra

A End to End Computer Vision Engine for Deep Learning Related Tasks

3 stars 0 forks source link

Add Functionality for Object Detection #1

Open harsh2ai opened 1 month ago

harsh2ai commented 1 month ago

Implement something like existing workflow for classification but for object detection which can be extended to tracking as well
Restructure the entire code base accordingly

Sakalya100 commented 1 month ago

Hi @harsh2ai. The project is great. I can work on this feature and add object detection support as well for this. Can you assign me this issue? I will figure out a checklist and share.

harsh2ai commented 1 month ago

@Sakalya100 could you share what do you have in your mind before I assign this issue to you a brief overview at least before you give me a detailed proposal , this library is like a long term help solution for a lot of folks in the industry/academia and kaggle to reduce their downtime on reduntant tasks and do quick poc

Sakalya100 commented 1 month ago

@harsh2ai Here is what I have on my mind:

Create a pipeline for Object detection in images
Add support for different huggingface models to be used for the task
Based on the response of different models, create bounding boxes and export the image locally for further use

Let me know if this works

harsh2ai commented 1 month ago

@Sakalya100 I am still figuring it out what model to use and what should be the ideal case for architechting the structure for library, can you instead add functionality which supports BLIP-2 for image labelling and finetuning it on a zero shot setting for classification purposes for now I might take a couple of more days to think of the issue in object detection terms

Sakalya100 commented 1 month ago

@harsh2ai Sure. I will take up and start working on the BLIP-2 support

harsh2ai commented 2 weeks ago

@Sakalya100 have you started working on this ?