-
i want to train it
hrlqq updated
6 years ago
-
下载了预训练模型,里面没有看到train_val.prototxt。查看了deploy.prototxt, train_val.prototxt跟example.prototxt配置不同吗?
-
Hi @liorshk
Thanks for sharing your code. It seems clean and well-written, however, I had problem having it converge.
I trained it on filtered version of MsCeleb with 5 Million images and 79…
rAm1n updated
5 years ago
-
@wy1iu
hi,
1 have two problems:
1)when i train the model with “type: SINGLE”,the loss can not converge,it is always 9.xx
2)then i use the model trained by 1),to train a new model,…
-
# 論文
ArcFace: Additive Angular Margin Loss for Deep Face Recognition
https://arxiv.org/pdf/1801.07698.pdf
2018/01/23
# 動機
レシピの画像検索について発表したときに,Face Recognitionで似たような手法(Triplet Loss)を利用しており,別の手法も…
-
Hi, I have tried first two step, the result seams good, FN does help converge a lot. But what do you mean by
> finetune lambda_min and lr
Do i need to adjust lambda and lr to tuning, if so wha…
-
Hello, ReDimNet is a very valuable work, and thank you very much for your open source sharing of the code.
I saw in the paper that using sphereface 2 as the loss function would yield better results…
-
Hi,
I have just evaluated your 20-layer CNN architecture caffemodel in Megaface.
The results are below:
Rank-1 Identification Accuracy with 1 Million Distractors, Set1 (FaceScrub )
77.6892% (75…
-
-
Hi, when I trained the sphereface-27 with my dataset, I always run into a problem:loss=87.33, and my prototxt as shown below:
layer {
name: "fc6-"
type: "MarginInnerProduct"
bottom: "fc5"
…