Marco Acerbis

> Yes, running full yolo or tinyYolo would be interesting. They update the Sipeed blog with a guide to train,convert and run MobileNet on the MaixPy board. http://blog.sipeed.com/p/680.html#more-680 Also here...

Thanks @AIWintermuteAI, I'll try to implement the object detector as well. I'm collecting some works about Transfer Learning, both developed with TF/Keras (Python) and in ML.NET(C#), in this repo: https://github.com/Ace95/transferlearning_AIDay...

Any suggestion to generate the dataset.txt file? How do you get the "truth boxes" on the training data? Thanks for the help!

I wrote this little python script that allows you to draw the "truth boxes" on the images and saves the cooordinates in dataset.txt. https://github.com/Ace95/tensorflow-yolov3/blob/master/Dataset_gen.py

> @Ace95 when ran your script, I got a error: cv2.error: OpenCV(4.1.0) /io/opencv/modules/highgui/src/window.cpp:352: error: (-215:Assertion failed) size.width>0 && size.height>0 in function 'imshow' > so,what your opencv verison？ Sorry for late...

> Hi, > > I use mobilenet v1, train and then convert to kmodel but there is error: > > Fatal: Invalid dataset. > System.ArgumentException: Invalid dataset. > at NnCase.Converter.Data.Dataset..ctor(String...