dsbyprateekg / yolov4 Goto Github PK

Hi, thanks so much for the blog and the repo.
I tried to follow every step in ipynb, however I get the following error when trying to load the saved model.
Any idea how to solve this issue?

Hi, I was having the followed error running your code:

TypeError Traceback (most recent call last)
in
7 for i in range(len(anchors)):
8 # decode the output of the network
----> 9 boxes += decode_netout(prediction[i][0], anchors[i], obj_thresh, input_h, input_w, len(anchors), scales_x_y[i])
10
11 print("nb boxes detected; ",len(boxes))

in decode_netout(netout, anchors, obj_thresh, net_h, net_w, anchors_nb, scales_x_y)
38
39 boxes = []
---> 40 netout[..., :2] = _sigmoid(netout[..., :2]) # x, y
41 netout[..., :2] = netout[..., :2]scales_x_y - 0.5(scales_x_y - 1.0) # scale x, y
42

TypeError: 'tensorflow.python.framework.ops.EagerTensor' object does not support item assignment

the "prediction" is the 'yhat' in your original notebook.

hi @dsbyprateekg
i have a yolov4 tiny model with 256X64 size. Is it possible to convert this into Tf with your repo?

I am using my own custom YOLOv4 weights file which i trained to detect 3 objects in the images.
When i am trying to execute cell where we have to call do_nms(boxes, nms_thresh) function then it is throwing error ZeroDivisionError: float division by zero . Here is a screenshot of the error

Hi there!
Great work, I like this workbook! I just study Yolo and I'd like to train this model on my own dataset. how can I du this ? which format of tensor I should provide to the model.
Thanks!

Can I load other weights to this model?