Francesco Ferroni

There are a number of conventional clustering algorithms that do not require a known number of cluster, such as DBSCAN. However, it might be tricky to establish your epsilon parameter...

Hi, Yes how I thought of doing it, and put that in the TODO comments. Feel free to fork and I can merge to the main branch! I just haven't...

Yea I guess that could work. For point n2. you could multiply by a fixed mask which masks that first column later on for other matrix operations, kind of like...

Right. The shape of the y where y = bboxutility.assign_boxes(y) is (7308,14). whereas (None, 20073, 14) is the size of the final layer if I use input_shape = (500,500,3). I...

@rykov8 Thanks, I use this script https://gist.github.com/codingPingjun/aa54be7993ca6b2d484cccf5a2c5c3d4 to generate new prior boxes with a new image_width, img_height. Probably the box_configs need changing too? @oarriaga Yes, but the objects I am...

Well, instinctively I would say that one needs to change the prior box fed into the bbox utility. Increasing the size of the input image for a fixed convolutipnal stack...

@ahrnbom Awesome thanks! Did you based the box_configs on the Cafe implementation?

If you have changed the initial convolutional stacks from a VGG to a residual type architecture, you simply need to ensure that the output convolved tensors are the same after...

Did you test the non-square modification? I ported the weights by adding the --fully_convolutional flag (`./yad2k.py yolo.cfg yolo.weights model_data/yolo.h5 --fully_convolutional`) so that it does not require known input size (although...

I tried the code on master, but also the two fixes in this thread, but it still gets stuck on "TRYING TO INITIALIZE" for my data. It's a series of...