hwac-dl / hwac_object_tracker Goto Github PK

hello. I want to replace the weights document to test another objects. However, it seems that the structure of the project's weights is not identical to that of the original tiny yolov2. How did you fix it?
It's very important for me.
I'm looking forward your reply.

this's the structure of weights95tuned.npy:
(152, 4)
(1168, 4)
(4640, 4)
(18496, 4)
(73856, 4)
(295168, 4)
(1180160, 4)
(1179904, 4)
(64250, 4)
this is the structure of the original tiny yolov2' weights:
(16, 3, 3, 3)
(16,)
(32, 3, 3, 16)
(32,)
(64, 3, 3, 32)
(64,)
(128, 3, 3, 64)
(128,)
(256, 3, 3, 128)
(256,)
(512, 3, 3, 256)
(512,)
(1024, 3, 3, 512)
(1024,)
(1024, 3, 3, 1024)
(1024,)
(125, 1, 1, 1024)
(125,)

Dear Sir,

Just wonder if this object detection repo supports multiple objects detection?
If not, is it possible for me to fetch all results from the last layer and get all the bounding boxes using pynq?

Thanks,

Brian

Can you please provide the testbench Verilog for Simulation? We want to studyeach block seperately.