Unable to convert tiny_yolo_v3 weights to .h5 file

Qidian213 / deep_sort_yolov3

Real-time Multi-person tracker using YOLO v3 and deep_sort with tensorflow

GNU General Public License v3.0

1.65k stars 593 forks source link

Unable to convert tiny_yolo_v3 weights to .h5 file #86

Open gaurav67890 opened 5 years ago

gaurav67890 commented 5 years ago

I am unable to convert tiny yolov3 weights to h5 file.

Evertime I run this command python convert.py yolov3-tiny.cfg yolov3-tiny.weights model_data/yolo.h5

I get this error Unsupported section header type: maxpool_0..

I am using tensorflow version 1.5.0 and keras version 2.2.2.

Does anybody has the saved .h5 file for tiny yolo v3 or know how to fix this error ? I need to run this code with fps of above 15. Right now it is running with 7fps on my gtx 1050ti.

bensloane commented 5 years ago

See https://github.com/qqwweee/keras-yolo3/commit/0f3567cc6e8c720e3cb7211670a5b60a1ae072b5 . Support for tiny-yolov3 architecture was enabled with an updated convert.py. That error is related to not having MaxPooling2D imported from keras as well as an accompanying parser for that layer type. Using the referenced convert.py file should help you solve this.

q823049029 commented 5 years ago

see https://github.com/qqwweee/keras-yolo3