Caffe Users - Google Groups

Groups

1–23 of 8727

0 selected

danie...@gmail.com2

8/7/18

Using MobileNet SSD training, loss always being 0

I solved this bug. It turns out the coordinates of the bbox in the annotation xml files cannot be

unread,

Using MobileNet SSD training, loss always being 0

I solved this bug. It turns out the coordinates of the bbox in the annotation xml files cannot be

8/7/18

3/7/18

Why I need very low learning rate in training SSD?

I used SSD for object detection. I trained 720 images with 720 x 720 width and height. Resize is set

unread,

Why I need very low learning rate in training SSD?

I used SSD for object detection. I trained 720 images with 720 x 720 width and height. Resize is set

3/7/18

Kostas, Pasc Peli3

3/12/18

High loss training fcn8s at once

Hi Pasc, It's been a while and I honestly can't remember what I did and unfortunately I don

unread,

High loss training fcn8s at once

Hi Pasc, It's been a while and I honestly can't remember what I did and unfortunately I don

3/12/18

11/9/16

solver type Adam worse than SGD

I use Dataset: cifar10 model: googlenet (inception v1) learning rate: fixed, 0.001(0~60000 iter),

unread,

solver type Adam worse than SGD

I use Dataset: cifar10 model: googlenet (inception v1) learning rate: fixed, 0.001(0~60000 iter),

11/9/16

Sharp Weapon, Lemma2

11/6/16

Training error and validation error are the same - zero accuracy during training

Hi, It seem you are using high lr_base, and opened the layers wight learning into high learning.

unread,

Training error and validation error are the same - zero accuracy during training

Hi, It seem you are using high lr_base, and opened the layers wight learning into high learning.

11/6/16

Sharp Weapon, Wilf Rosenbaum4

11/2/16

Zero accuracy training a neural network using caffe

Oh, ok, in that case the best thing you should check is

unread,

Zero accuracy training a neural network using caffe

Oh, ok, in that case the best thing you should check is

11/2/16

Suyog Trivedi, Ketil Malde5

10/14/16

Caffe significance of Validation (test) loss and Train (loss)

Thanks for your help. I ran the training for 15k iterations. I am getting the output curve as below.

unread,

Caffe significance of Validation (test) loss and Train (loss)

Thanks for your help. I ran the training for 15k iterations. I am getting the output curve as below.

10/14/16

7/22/16

Finetuning: Stopping some layers from learning via "lr_mult: 0" doesn't seem to work

Or maybe it doesn't use the former net, after all: https://groups.google.com/forum/#!searchin/

unread,

feature-extraction

Finetuning: Stopping some layers from learning via "lr_mult: 0" doesn't seem to work

Or maybe it doesn't use the former net, after all: https://groups.google.com/forum/#!searchin/

7/22/16

Raúl Gombru, Evan Shelhamer3

7/28/16

Variance in optimum learing rate value to fine-tune FCN in different frameworks

Thank you for your answer Evan! It has been really helpful. So I understand that the choice of

unread,

fullyConvolutional

Variance in optimum learing rate value to fine-tune FCN in different frameworks

Thank you for your answer Evan! It has been really helpful. So I understand that the choice of

7/28/16

Jalen Hawkins, Daniel Moodie4

6/24/16

okay so i found a few slight mechanical errors such as the name of some of the pictures not matching

unread,

okay so i found a few slight mechanical errors such as the name of some of the pictures not matching

6/24/16

dpvo, Norman he2

6/10/16

loss suddenly jumps as fine-tuning GoogleNet

It is overfitting after the 119040th iteration. On Sunday, May 15, 2016 at 12:49:43 PM UTC-7, dpvo

unread,

loss suddenly jumps as fine-tuning GoogleNet

It is overfitting after the 119040th iteration. On Sunday, May 15, 2016 at 12:49:43 PM UTC-7, dpvo

6/10/16

3/23/16

Including `phase` and `blob_lr` in net prototxt from Python

This code: from caffe import layers as L from caffe import params as P def lenet(lmdb, batch_size): #

unread,

Including `phase` and `blob_lr` in net prototxt from Python

This code: from caffe import layers as L from caffe import params as P def lenet(lmdb, batch_size): #

3/23/16

Jan C Peters, Evan Shelhamer3

10/30/15

Why learning rate in AdaDelta?

Thanks for the answers, Evan! Am Donnerstag, 29. Oktober 2015 22:00:43 UTC+1 schrieb Evan Shelhamer:

unread,

Why learning rate in AdaDelta?

Thanks for the answers, Evan! Am Donnerstag, 29. Oktober 2015 22:00:43 UTC+1 schrieb Evan Shelhamer:

10/30/15

Harshit Laddha, … Miguel Garcia3

4/1/16

Implementing GoogLenet for Devanagri (Hindi) character classification.

Try centering your training images El sábado, 22 de agosto de 2015, 2:06:00 (UTC-5), Harshit Laddha

unread,

Implementing GoogLenet for Devanagri (Hindi) character classification.

Try centering your training images El sábado, 22 de agosto de 2015, 2:06:00 (UTC-5), Harshit Laddha

4/1/16

Suraj Srinivas, Evan Shelhamer3

6/2/15

Accuracy changes even when learning rate is set to zero

Thanks for your response! I am not entirely sure how it works. So if I delete the solverstate file

unread,

Accuracy changes even when learning rate is set to zero

Thanks for your response! I am not entirely sure how it works. So if I delete the solverstate file

6/2/15

3/4/15

publish a tools [huabot-brain] for easy using caffe

I write a web base front tool for easy use caffe please see https://github.com/Lupino/huabot-brain

unread,

publish a tools [huabot-brain] for easy using caffe

I write a web base front tool for easy use caffe please see https://github.com/Lupino/huabot-brain

3/4/15

Xiao Yang, … Evan Shelhamer11

2/24/15

Get "loss=nan" info at the very beginning; even when setting base_lr=0

Thanks! That helps a lot, since right now I am trying to have a look of the source code Besides,

unread,

Get "loss=nan" info at the very beginning; even when setting base_lr=0

Thanks! That helps a lot, since right now I am trying to have a look of the source code Besides,

2/24/15

Zhongyu Lou, … ada...@ucr.edu4

10/24/16

What's the rule of gamma and weight_decay is solver.prototxt

From some old discussions (link1, link2) I got the idea that 'weight_decay' parameter is the

unread,

What's the rule of gamma and weight_decay is solver.prototxt

From some old discussions (link1, link2) I got the idea that 'weight_decay' parameter is the

10/24/16

Aleš Žurek, … Alex Orloff6

9/3/16

Nudity detection

Hi Sergey! Thank you for your reply So, can you please share your solution? BTW does your model make

unread,

Nudity detection

Hi Sergey! Thank you for your reply So, can you please share your solution? BTW does your model make

9/3/16

10/24/14

adagrad and learning rates

How does the learning rate policy affect adagrad? Should we just set a fixed learning rate and let

unread,

adagrad and learning rates

How does the learning rate policy affect adagrad? Should we just set a fixed learning rate and let

10/24/14

Ankit Arya, … Evan Shelhamer4

10/5/14

Example: Fine tuning flickr dataset

Thanks Yangqing and Evan, that was helpful ! I get the point now. Sorry if it was a naive question,

unread,

Example: Fine tuning flickr dataset

Thanks Yangqing and Evan, that was helpful ! I get the point now. Sorry if it was a naive question,

10/5/14

9/15/14

Why the model was changed even if setting blobs_lr = 0?

Greetings, Thanks for developing such an amazing tool for deep learning. I'm using Caffe and

unread,

Why the model was changed even if setting blobs_lr = 0?

Greetings, Thanks for developing such an amazing tool for deep learning. I'm using Caffe and

9/15/14

9/12/14

Is there a way to change the learning rate while training?

I know I can change the learning rate and start from the last snapshot. But I was wondering if it is

unread,

Is there a way to change the learning rate while training?

I know I can change the learning rate and start from the last snapshot. But I was wondering if it is

9/12/14

Search

Clear search

Close search

Google apps

Main menu