I’m new to machine learning especially to the subject of Object Detection. If you don’t mind can you please share your experience with me and name a good package to start with? I’m really confused about the different packages.
Your help will be appreciated.
If you wish to use pretrained networks, have a look at these YOLO implementations which can load weights files:
- Yolov2tiny working. Full YoloV2 working in my fork of the repo. Here you can see the network architecture quite easily. This is based on Knet.
- YoloV2 and YoloV3 trained on multiple datasets. The network architecture is harder to read, but it has the big advantage of autogenerating the networks based on cfg files. This is based on Flux.
I know @ianshmean had some attempt of writing the loss function in order to train the the network from scratch in Julia, but I don’t know the current status. I am also trying to port the loss function from Tensorflow implementations, but haven’t finished it yet.