Comments (11)
I was able to train it on a GTX980, 4GB.
However, I had to change the batch size to 2. Unfortunately the results are not really well.
from caffe-segnet.
I am surprising, that some have problems with 4 GB. I have only changed the batch size.
Somewhere I read that cudnn version 2 is required. Actually I use version 3 (ver. 3.0.07, CUDA ver. 7.5). I could imagine that this is the reason.
from caffe-segnet.
Just tried it on my laptop (Nvidia GTX970M 4GB) and had the exact same result as issue #21, I am unable to train. This issue was never resolved so I am not sure how to move forward.
Is 4GB really not enough or is something else causing the problem?
from caffe-segnet.
Thanks! I've been able to run SegNet on my laptop (GTX960M, 4gb) and a Jetson TK1 and Jetson TX1. As some advice, to reduce memory you can play with things like: input resolution, batch size, network width and depth.
from caffe-segnet.
That is promising! I was hoping to deploy on the TX1.
Would you happen to have the prototext files for training on the 960M or did you only deploy to the laptop? I have tried reducing the images to half their resolution and batch size is only 1 but I am still having memory issues. Exactly the same as issue #21
Also I was wondering what kind of performance you got on the TX1. How much lag is there with live video?
from caffe-segnet.
I must correct myself. The results are after 40,000 iterations quite well. Before that I only tested with 10,000 iterations.
from caffe-segnet.
Thanks for sharing. Good to know that 4GB can train with good results. Could you tell me what modifications you made in order to train on 4GB? I have not had any success with this. Was changing the batch size the only thing you changed!? I have tried with a batch size of 1 and still ran out of memory. Perhaps something wrong with my environment...
from caffe-segnet.
I'm also encounter memory problem.
I followed Segnet tutorial (https://githucom/alexgkendall/SegNet-Tutorial) on Jetson TX1 and ran webcam_demo.py.
It crashes when it tried to do forward propagation of the network
out = net.forward_all(data=input_image)
and the error is
I0611 11:46:34.369215 19670 net.cpp:247] Network initialization done.
I0611 11:46:34.369243 19670 net.cpp:248] Memory required for data: 1065139200
Here is videos
Grabbed camera frame in 7.64513015747 ms
Resized image in 126.845121384 ms
Killed
I cleaned the cache and ran webcam_demo.py, but it doesn't work.
Is there any suggestion?
My memory status is
total used free shared buffers cached
Mem: 3853 1735 2118 40 24 282
-/+ buffers/cache: 1428 2425
Swap: 0 0 0
from caffe-segnet.
Thanks again Timo-hab. I had been told that I needed CUDA v2 for gpu acceleration but wasn't certain if this would fix the memory issue. Seems that this is the case then. Will try the old version...
from caffe-segnet.
Does it worth to fight to install segnet on a laptop with a nvidia gt740M ? I plan to train on greyscaled images not larger than 200x200.
~/App/samples/bin/x86_64/linux/release$ nvidia-smi
Mon Jul 18 16:21:43 2016
+------------------------------------------------------+
| NVIDIA-SMI 352.63 Driver Version: 352.63 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce GT 740M Off | 0000:01:00.0 N/A | N/A |
| N/A 53C P0 N/A / N/A | 5MiB / 2047MiB | N/A Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 Not Supported |
+-----------------------------------------------------------------------------+
from caffe-segnet.
I also have memory problem on Jetson TX1.
here is the error message:
I0324 14:17:41.216102 28333 net.cpp:247] Network initialization done.
I0324 14:17:41.216142 28333 net.cpp:248] Memory required for data: 1065139200
Grabbed camera frame in 2.96902656555 ms
Resized image in 7.36689567566 ms
Killed
anyone has idea to solve it?
Thanks!
from caffe-segnet.
Related Issues (20)
- About Run In C++
- Separate input concatenation in segnet
- Caffe2 caffe-segnet HOT 2
- make runtest fail HOT 4
- Reducing the noise in a image
- Error occurred when make resource code
- Which file does caffe execute when implementing convolutional operation?
- CUDNN version
- Caffe-SegNet installation on Windows
- All indices of the max locations will be stored and passed to the decoder during pooling?
- Hey - it doesn't make sense to train segnet on a single class. In this situation, segnet will simply learn to predict every pixel as this class, and you will see loss=0. HOT 1
- Help for implementing SegNet for predicting Single class
- Check failed: error == cudaSuccess (77 vs. 0) an illegal memory access was encountered HOT 2
- make problem
- make error
- nvcc fatal: redefinition of argument 'compiler-bindir'
- what is the softmax uncertainty in Bayesian Segnet paper?
- changing the number of classes in segnet produce error: cudaSuccess (700 vs. 0) an illegal memory access was encountered. HOT 1
- Try-demo issues about website http://mi.eng.cam.ac.uk/projects/segnet/demo.php
- why last prediction layer is 3x3 convolution??
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from caffe-segnet.