avanetten / yolt Goto Github PK

可以不按照提供的python格式，用yolo本来的训练方式训练吗？

The terminal shows the number of iteration and doesn't go on .
This program doesn't stand the cuda memory.

Hi @ @avanetten ，
Your project is such cool. I am interested in them. I want to improve them and cite your paper. But, can you share more codes and more datas so that I can test them and train them?

Bests,

Region Avg IOU: nan, Class: nan, Obj: -nan, No Obj: -nan, Avg Recall: 0.000000, count: 16
Region Avg IOU: nan, Class: nan, Obj: -nan, No Obj: -nan, Avg Recall: 0.000000, count: 8
Region Avg IOU: nan, Class: nan, Obj: -nan, No Obj: -nan, Avg Recall: 0.000000, count: 18
Region Avg IOU: nan, Class: nan, Obj: -nan, No Obj: -nan, Avg Recall: 0.000000, count: 12
Region Avg IOU: nan, Class: nan, Obj: -nan, No Obj: -nan, Avg Recall: 0.000000, count: 9

Gradient explosion occurred while training cowc data sets. What should I do?

Why I got an empty prediction file and no predicted png image using validation mode? I do have ever get those results when I tried another model using different dataset.

Hello, thank you very much for sharing! When I was running yolt2.py, I met such a mistake: No such file or directory:'/raid/local/src/yolt2/', what's the reason for it, can you help me to solve it?

Hi,@avanetten
When installing gdal2.1.0 with conda,why is the display:
CondaRuntimeError: Runtime error: RuntimeError: Runtime error: Could not open u'/home/cp/anaconda2/pkgs/geos-3.5.0-0.tar.bz2.part' for writing (HTTPSConnectionPool(host='repo.continuum.io', port=443): Read timed out.).

The gsd of the original cowc dataset is 0.15,how can I train the cowc with 0.3gsd?

Please guide us through the commands needed for running yolt 2.0 as we run darknet @avanetten

when specifying the dataset, are x, y, width, height to be specified in absolute pixes, between 0 and 1 relative to the size of the whole image, or in some other way?

Thanks!

the data list show the data path is "/raid/local/src/yolt2/training_datasets/rio_planes_312/training_data/images/slice_airplane_rgb0_0_2490.jpg"
but I can't find in the dataset folder, Do i miss something?

Hi @avanetten ,

Thanks for your work. I have read some articles talking about your paper and since now I am working on traffic-sign detection (though it is a much easier problems than yours), it will be great if we can make some discussion on this topic.

From my perspective, your model generally based on YOLO2, and the improvement are mainly on:

1. crop the big picture to several small one, and than detect each
2. modify the network so the last feature map will be 26^2 but not 13^2
   Am I right?
   Actually, I am now using YOLO3 to detect traffic sign. The picture in my dataset is 2048^2, the size of ground truth boxes vary from 10 pixels to around 100 pixels. Like you, I also do random crop to make the input size become 512. Unfortunately, the result is pretty bad. I have also tried other input size such as 1024, but the result seems similar.

So, have you tried using your strategies in YOLO3? And, any other structures or parameters modification do you think will do good to small object detection?

Many thanks.

Hi, thank you for your sharing firstly. I have deployed your code in docker but when i try to compile the Makefile it makes mistake like that:
make: *** No rule to make target 'obj/yolt.o', needed by 'darknet'. Stop

And I note that the line 24 in Makefile "COMMON=-I/Users/avanetten/anaconda/pkgs/libgdal-2.1.0-0/include" , I just use python2.7 of system and already add PPA for "pip install gdal", but still make failed.
So how should I fix it?

Hi, how could I split large images for training like yolt does it for validation and keep bounding boxes when splitting?

Dear @avanetten
Can you please provide the pretrained model weights as well as the training code in python/any other deep learning framework.
It would be really helpful
Thanks

Hi i am trying to run yolt but unable to find YOLT configuration file in the repo please help
and where are the .weights file

Hi there!
I want to convert yolt to a matlab tool, but I can't find the passthrough function in the code or cfg files.
Maybe this is because I don't know the bottom principle of your project.

Could you please tell me how you achieve this passthrough function?

Can the dataset used by you for detecting ships and planes be provided in the public domain. It would be really helpful.
Thanks @avanetten

problem when I run yolt2.py

nvidia-docker build -t yolt yolt_docker_name /path_to_yolt/docker

In following script where is the ave_13x13.cfg file

python yolt2.py
--mode train
--outname 3class_boat_plane
--object_labels_str boat,boat_harbor,airplane
-- cfg_file ave_13x13.cfg
--nbands 3
--train_images_list_file boat_airplane_all.txt
--single_gpu_machine 0
--keep_valid_slices False
--max_batches 60000
--gpu 0