Konstantin Shmelkov

Hello, I used the official Matlab implementation of [edgeboxes](https://github.com/pdollar/edges) to generate them. Parameters are in the file `compute_edgeboxes.m`. I guess there was also code to aggregate all matlab-generated proposals in...

Thank you for the explanation. Probably I didn't really explain my question, rereading it now I realize that it isn't clear at all. I didn't ask how to implement _this...

@ebenolson Parameters become disconnected from layers. Suppose you want to pass two sets of parameters: number of filters and kernel size. Either they are two separate lists like `[64, 128,...

Actually I don't have an exact idea in mind how to decouple parametrisation and network topology. I was thinking about proper dispatching arguments in corresponding layers' `**kwargs` (in order to...

Can't it be emulated via `TransformerLayer`? IIRC, it should do the trick if you transform bbox coordinates into affine transform parameters.

Fair point. Somehow I missed that it is called pooling for reason. Anyway I am messing around faster rcnn and I almost finished implementation of ROI 'pooling' via `TransformerLayer`. I...

I'd like to apply CTC to some OCR tasks. May be we can imagine a synthetical text dataset? It would be much easier to visualize labelling on images. If this...

@rakeshvar thank you. Actually I used pretty successfully the code of Mohammad Pezeshki and Philemon Brakel (the base of @skaae Lasagne implementation). I have seen yours, but didn't understand how...

`FC_WEIGHT_DECAY`. Which already exists, but just isn't used. For equivalent code in conv layers `CONV_WEIGHT_DECAY` is used.

Honestly I can't understand the purpose of that function. I couldn't find any equivalent in the original code. As far as I understand, `pseudo_cost` computes gradients manually instead of relying...