- See in Colab Notebook
- Open in Jupyter
-
Imp Files Modified in Yolov4 + deepsort
- INFO.md: contains all the referenced and info about he model used for Yolov4 and deep sort
- object_tracker.py : Implements yolov4 + deepsort and stores the segmented image of each object in a folder
- algoa.py : Implements yolov4 + deepsort + DeepLogo for realtime detection according to the algo A as given in algo.md files
- Outputs a Dict contains object , their ids and detected logo
- algob.py : Implements yolov4 + deepsort + DeepLogo for realtime detection according to the algo B given in the algo.md files
- Ouputs a dict containing object , their ids and detected logo
-
Imp Files in Deep Logo
- single_inf.py: contains all the function that is used with the object tracker files to detect logo .. (Stands for single inference!)
- annotations/: contains the dataset info (train,test classes distribution).Also has a parser to read through the annotation files
- imgs/: contains the custom dataset for testing the model
- MODEL-PERFORMARMANCE.md : Contains the info about the testing process of the model and about the results
- Testing Process and Notebook Demonstrating Working of Logo Detection
- Read algo.md
- Open Graph with links
- Might Require to create an account
- algoa-tech
- Open Graph with links
- Might Require to create an account
- algob-tech
- Train a single model to give object and the logo both?
- Add more diverse images .. as images of some classes like apple are not diverse enough
- Test More as to what qualilty of image yeilds best detection results .. then maybe apply some processing filtering techniques to make detection accurate (hence also faster)
- Research on Sequence based methods to process video?