-
Notifications
You must be signed in to change notification settings - Fork 8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
KITTI Results #4454
Comments
@CuongNguyen218 Hi,
|
@AlexeyAB , Hi, |
Try to use this repo: https://github.com/AlexeyAB/darknet I just added |
@AlexeyAB ,
I see it is an unreasonable result follow gaussian paper:
|
|
@AlexeyAB , |
@AlexeyAB , |
There isn't mAP that is based on GIoU instead of IoU. Try to ask it in GIoU-repository.
Just Bilateral-blurring will be used for 50% of images even if you didn't enable it. |
@AlexeyAB , |
GIoU just replaces NLL loss for b-box. |
@AlexeyAB , I retrained KITTI dataset with your new code and your valid code but something when wrong with car class. The trend is Easy > Moderate > Hard but my result what I get from your code is not follow that. |
Also you can try to set in your cfg-file
Then train usin AlexeyAB repo, and valid by using Jwchoi repo. |
@AlexeyAB , Thanks for your support. and i want to find commit of bag of freebie. Can you find it for me? |
Serach commit with MixUp. or just use
|
Yes. |
But you should know, that mixup=1 improves accuracy only for Classifier. (and the use Classifier pre-trained weights file for Detector). For training Detector is recommended to use |
okay, I got it. Thanks you. I want to try it because i see it improve performance of yolov3 in gluoncv +5% on COCO. |
BoF (Bag of Freebies) includes 5 features - there are implemented 5 of 5 features in this repository, read: #3272 (comment) There is implemented
Paper: https://arxiv.org/pdf/1902.04103.pdf To get To get another +2.0% you should use: mixup=1, SGDR (cosine lr), random=1, and GPU-processing + CPU-RAM - it takes a lot of time to training. There are better approaches than MixUP:
|
@AlexeyAB , |
What width= and height= do you use in both cfg-files? |
Hi @AlexeyAB , |
Did you set Can you attach both cfg-files? There is also different network resize approaches:
|
Yes. |
Did you train both models with the same batch and subdivisions?
Try to change these lines: Lines 169 to 171 in 318919e
To these int dim = (rand() % 10 + 10) * 32;
if (get_current_batch(net)+200 > net->max_batches) dim = 608;
int dim_w = dim;
int dim_h = dim; recompile and train without letter_box=1 |
Yes, I set both are Batch=64 and sub =16 |
Use Also removes these lines: Lines 173 to 177 in 318919e
|
@AlexeyAB |
Use |
@AlexeyAB , thank you it work. As I feel, training speed is faster. |
@AlexeyAB , |
Hi, if you are using |
Use Can be: voc, coco, imagenet, kitty, bdd: Lines 662 to 708 in 49bff0e
|
If you can't get Kitti-evalutation, also you can try to use BDD-evalutation methond
|
Hi Alexey,
|
@AlexeyAB ,
Hi Alex. Thanks for your amazing code. But I have a trouble when Training your repo with KITTI Dataset.
When I use your repo and trained around 50k iter, I get a bad result with car AP detection when test with KITTI Devkit. Details in image below.
And my result at final iter with your repo below
And chart image of training process below:
I see a gap between the result of car mAP when I use your repo and GaussianYolo repo is large. Notice that I don't use Gaussian Yolo Layer. Result from Gaussian Yolo repo in the image below
Can you explain it for me ?
This is my config file
yolov3-kitti.txt
:
And this is the number of box of KITTI dataset in training set and test set:
The text was updated successfully, but these errors were encountered: