Re-trained weight of feature extractor model for vehicle instead of MARS

Akhtar303 / Vehicle-Detection-and-Tracking-Usig-YOLO-and-Deep-Sort-with-Keras-and-Tensorflow

Vehicle-Detection and Tracking Using YOLO and Deep Sort with Tensorflow

MIT License

34 stars 12 forks source link

Re-trained weight of feature extractor model for vehicle instead of MARS #1

Open sangnguyenz opened 5 years ago

sangnguyenz commented 5 years ago

Thanks for your implementation for vehicle detection and tracking by YOLO&DeepSort. I had a question that why you still use the pre-trained weight on MARS dataset (pedestrian dataset) to compare appearance between vehicles?

Do you think about training on vehicle dataset with proposed network (cosine metric learning)?

Let me know if you have any improvement. Thank you!

Akhtar303 commented 5 years ago

@sangnguyenz yes you can try it.

sangnguyenz commented 5 years ago

I can see in the model_data folder contain frozen_inference_graph.pb. Can you tell me if this weight was used for what? Thank you for your instant reply.

quocnhat commented 4 years ago

@sangnguyenz have you retrained the weight to extract vehicle embedding? I want to try but I have no idea how to do it. please give me some hinds, thank you

sangnguyenz commented 4 years ago

@quocnhat Currently I haven't continued to work on this project, so that I haven't tried anything since September. Sorry for can not help you with this.