These are chat archives for
Sign in to start talking
Deep Learning API and Server
Oct 23 2017 17:56
I've joined on gitter here to try to understand the nuances of what's going on with deep detect
I have managed to run cchadowitz-pf's extract_layer_nsfw to do a flattened output of an image
Oct 23 2017 18:23
can confirm that the difference comes from the input images (as read by PIL vs read from OpenCV) ?
Oct 23 2017 18:33
I have found the output of my vgg model and the output of the yahoo model matches up.
This is if I am interpreting the results of the extract_layer_nsfw.py correctly when inputting the same image but inserting a different model than the yahoo_nsfw.
sudo docker run --volume=$(pwd):/workspace caffe:cpu python ./extract_layer_nsfw.py --extract-layer data --layer-file testvgg --model_def vgg_model/deploy.prototxt --pretrained_model vgg_model/snapshot_iter_31410.caffemodel test_image1.jpg
Oct 23 2017 21:38
What do you mean by 'the output of my vgg model and the output of the yahoo model matches up' exactly?
Oct 23 2017 21:53
Scratch that thought, I'm still working on it.
I hope tomorrow we can go over some details of your testing
Oct 23 2017 23:30