Alexander Lazarev
Alexander Lazarev
@eldar Thank you! BTW what's the test results does it have in comparison to ResNet-152?
Oh, yes I've forgotten about that table. Thanks!
@energabos did you get `out = net.forward_all(data=input_image)` working? for me it raises an exception `Input blob arguments do not match net inputs`
2.7.6, but in your README mentioned that it should run on 2.7
I guess it can be replaced with `*`
I'm afraid np.dot() is very inefficient in this case. It hangs my i7 for a minute and then a process is killed.
So I've tried both `*` and `@`. The same speed and results. I think `*` works because of numpy's broadcasting
guys, check out my fork https://github.com/Arsey/deepsaldet where you can avoid all these issues using docker: 1. build image `sudo nvidia-docker build --no-cache -t deepsaldet .` 2. run container `sudo nvidia-docker...
@qingliucs checkout my fork https://github.com/Arsey/deepsaldet. It has docker so you can build it automatically with docker and execute: 1. build image `sudo nvidia-docker build --no-cache -t deepsaldet .` 2. run...
I would recommend you to read the [paper](https://arxiv.org/pdf/1703.02529.pdf) more patiently. NoScope uses YOLO for detection but does it in a very efficient way.