This repo will contain detailed and more visualized tutorials(loss and results) on writing custom loss functions. Scope will be ranging from basic neural networks to RL. it will also explain about existing loss function with SOTA results.
(We will have visuvalized graphs, explanation for why we used a particular loss function, explanations for why our custom loss function is working well. Also corresponding visuvalizable results will be shared.) Following will be covered:
Classification-
- Normal CNN layers networks
- Resnet, Densenet
Detection
- Yolo
- SSD, Mobilenet
Tracking
- Deepsort+Yolo
Others
- 4 GAN Models
- 1 R-CNN model
RL - Single Agent
- DQN Model
- DDPG model
- TD3 model
RL - Multi Agent
- DQN Model
- DDPG model
- TD3 model
In case if above is finished we would proceed with building custom RL - Single agent and RL-Multi agent environments in next repo.
Completion of this would take a person from Beginer level DL programmer to intermediate level DL and RL programmer for sure.
Time lines:
Estimated start of project: After completeion of GANs repo
Estimated duration - 4 months to 5 months