Model not giving expected results after conversion from Darknet to Caffe

[miquelmarti]

After fixing a typo in the code (I guess) in create_yolo_prototxt.py as it says Relu but caffe seems to want ReLU, I could run both this script and create a prototxt which seems fine and create_yolo_caffemodel.py which created a .caffemodel file without errors. I wonder if this was using an old version of caffe otherwise I cannot understand how could it work for others...

However, running it with a sample image taken from ImageNet with standard caffe python interface classify.py script the result is not good. All classes get similar scores while the image I used gets a 0.99 score for the correct class with AlexNet. Anyone had the same problem? Is there anything I should be aware of when using create_yolo_caffemodel.py?

[michaelholm-ce]

I am having the same problem. I converted the darknet reference model cfg file to a caffe prototxt file and could not get it to converge on a standard dataset, in spite of trying a full range of learning rates. Any advice here?

[baristahell]

Is it related to this comment ?
I just can't convert my yolo.weights, same issue than number 7, and it might be related..

[miquelmarti]

Sorry guys I couldn't find a solution so I just moved on. If you do please share!

[michaelholm-ce]

Now that I've played around with the repo enough, I can simplify my problem even further.

I can't train a model in caffe on a known dataset using any of the prototxt files included in this repo. See my question #21. If anyone else can, I would appreciate hearing what hyper parameters you used or changes you made to get a model to learn.