Skip to content

Instantly share code, notes, and snippets.

View walkoncross's full-sized avatar

James Zhao walkoncross

95 followers · 576 following
View GitHub Profile
@walkoncross
walkoncross / FACE-SE-ResNeXt-101-train-L2Softmax.prototxt
Last active December 18, 2017 00:00
caffe-face train-val prototxt, netscope:http://ethereon.github.io/netscope/#/gist/cb2b1c90f0646da623a79c84bf9171d6
name: "FACE-SE-ResNeXt-101 (32 x 4d)"
# mean_value: 104, 117, 123
#layer {
# name: "data"
# type: "Input"
# top: "data"
# input_param: {
# shape: {
# dim: 1
# dim: 3
@walkoncross
walkoncross / coco-loss-face.prototxt
Last active December 16, 2017 19:26
network from coco_loss for face verification, netscope: http://ethereon.github.io/netscope/#/gist/7e737a9f3d00c6bd69d13fdd70069148
#final size:235
#det-ver:rsa-res-v3
#align-ver:affine-v1
#crop:110
#offsety:12
#batchsize:640
name: "IRv1"
mem_param {
optimize_train: false
@walkoncross
walkoncross / caffe-face-center-loss-trainval.prototxt
Last active December 15, 2017 22:52
caffe-face train-val prototxt, netscope:http://ethereon.github.io/netscope/#/gist/22f4f60d810d9591a82625288e2eda1c
name: "Face-ResNet-Center-loss"
layer {
name: "data"
type: "ImageData"
top: "data"
top: "label"
include {
phase: TRAIN
}
transform_param {
@walkoncross
walkoncross / caffe-face-center-loss-deploy.prototxt
Last active December 15, 2017 22:50
caffe-face deploy prototxt, netscope:http://ethereon.github.io/netscope/#/gist/8fed3c5d2db502cddfa8e9a1650883eb
# center loss face
# https://github.com/ydwen/caffe-face/blob/caffe-face/face_example/face_deploy.prototxt
name: "caffe-face-center-loss"
input: "data"
input_dim: 1
input_dim: 3
input_dim: 112
input_dim: 96
layer {
name: "conv1a"
@walkoncross
walkoncross / SE-ResNeXt-50.prototxt
Created December 14, 2017 16:27
SE-ResNeXt-50.prototxt, netscope: http://ethereon.github.io/netscope/#/gist/5c91ed8333d88877c077afabe1b1df42
name: "SE-ResNeXt-50 (32 x 4d)"
# mean_value: 104, 117, 123
layer {
name: "data"
type: "Input"
top: "data"
input_param: {
shape: {
dim: 1
dim: 3
@walkoncross
walkoncross / SE-ResNeXt-101.prototxt
Created December 14, 2017 16:25
SE-ResNeXt-101.prototxt, netscope: http://ethereon.github.io/netscope/#/gist/f5f4ba6a8950cf160aa6a48ebdf3e7a5
name: "SE-ResNeXt-101 (32 x 4d)"
# mean_value: 104, 117, 123
layer {
name: "data"
type: "Input"
top: "data"
input_param: {
shape: {
dim: 1
dim: 3
@walkoncross
walkoncross / ResNet-50-deploy-112x96.prototxt
Last active November 11, 2017 17:42
ResNet-50-deploy.prototxt from Kaiming He's repo, netscope: http://ethereon.github.io/netscope/#/gist/f251d13bc2d8f249a480b033b2023941
name: "ResNet-50"
input: "data"
input_dim: 1
input_dim: 3
input_dim: 112
input_dim: 96
layer {
bottom: "data"
top: "conv1"
@walkoncross
walkoncross / deploy.prototxt
Created November 11, 2017 13:25 — forked from tzutalin/deploy.prototxt
Network In Network
name: "nin_imagenet"
input: "data"
input_shape {
dim: 10
dim: 3
dim: 224
dim: 224
}
layers {
bottom: "data"
@walkoncross
walkoncross / sphereface_20_model_add_bn_v2.prototxt
Last active November 11, 2017 02:03
http://ethereon.github.io/netscope/#/gist/ac7a8977ade92a920b036dfe0200a90e
name: "BN-SpherefaceNet-20"
layer {
name: "data"
type: "ImageData"
top: "data"
top: "label"
transform_param {
mean_value: 127.5
mean_value: 127.5
@walkoncross
walkoncross / resnet_50_train_val.prototxt
Last active November 11, 2017 01:26
resnet_50_train_val.prototxt; Netscope: http://ethereon.github.io/netscope/#/gist/4156bda88b9a59906fd8e0d25424187a; from (https://github.com/beniz/deepdetect/blob/master/templates/caffe/resnet_50/resnet_50.prototxt)
name: "ResNet-50"
layer {
name: "resnet_50"
type: "Data"
top: "data"
top: "label"
include {
phase: TRAIN
}
transform_param {