xView.yaml label category mapping with YOLO has problem with id=75

[geobao]

Search before asking

• [X] I have searched the YOLOv5 issues and found no similar bug report.

YOLOv5 Component

Other

Bug

There is a yaml that is specific for the dataset of xView. (xView.yaml). In this file there is a python script to convert the original labels to YOLO-suitable labels. The original labels of xView contain ids contained 11-94. There are a total of 60 classes. Therefore in YOLO we want ids from 0 to 59 without gaps. There is a list that maps ids from the original id_classes of the dataset to the new ids that will be used in YOLO. However, there exists an id=75 in the original dataset. This id is not mapped to a correct YOLO id but it is assigned a -1. This will lead to mark all the images with this class as corrupted and they will be ignored in the training.

The mapping occurs in this line of code: https://github.com/ultralytics/yolov5/blob/24bea5e4b7ad369753f45e93b736d9205c37d20a/data/xView.yaml#L59

Environment

No response

Minimal Reproducible Example

To explore the original class ids from the dataset, the original labels can be loaded in a Dataframe like that: Download manually from https://challenge.xviewdataset.org current link for training labels > here

import gepandas as gpd
gdf = gpd.read_file( "./xView_train.geojson" )

Additional

No response

Are you willing to submit a PR?

• [ ] Yes I'd like to help by submitting a PR!

[github-actions[bot]]

👋 Hello @geobao, thank you for your interest in YOLOv5 🚀! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

If this is a 🐛 Bug Report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset images, training logs, screenshots, and a public link to online W&B logging if available.

For business inquiries or professional support requests please visit https://ultralytics.com or email Glenn Jocher at glenn.jocher@ultralytics.com.

Requirements

Python>=3.6.0 with all requirements.txt installed including PyTorch>=1.7. To get started:

$ git clone https://github.com/ultralytics/yolov5
$ cd yolov5
$ pip install -r requirements.txt

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Google Colab and Kaggle notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training (train.py), validation (val.py), inference (detect.py) and export (export.py) on MacOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

@geobao hi, thank you for your suggestion on how to improve YOLOv5 🚀!

The fastest and easiest way to incorporate your ideas into the official codebase is to submit a Pull Request (PR) implementing your idea, and if applicable providing before and after profiling/inference/training results to help us understand the improvement your feature provides. This allows us to directly see the changes in the code and to understand how they affect workflows and performance.

Please see our ✅ Contributing Guide to get started.

About this specific topic, training works correctly for us on xView, we are not able to reproduce any issues with corrupted images:

python train.py --data xView.yaml

train: weights=yolov5s.pt, cfg=, data=xView.yaml, hyp=data/hyps/hyp.scratch.yaml, epochs=300, batch_size=16, imgsz=640, rect=False, resume=False, nosave=False, noval=False, noautoanchor=False, evolve=None, bucket=, cache=None, image_weights=False, device=, multi_scale=False, single_cls=False, adam=False, sync_bn=False, workers=8, project=runs/train, name=exp, exist_ok=False, quad=False, linear_lr=False, label_smoothing=0.0, patience=100, freeze=0, save_period=-1, local_rank=-1, entity=None, upload_dataset=False, bbox_interval=-1, artifact_alias=latest
github: up to date with https://github.com/ultralytics/yolov5 ✅
YOLOv5 🚀 v6.0-45-g042f02f torch 1.10.0 CPU

hyperparameters: lr0=0.01, lrf=0.1, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=0.05, cls=0.5, cls_pw=1.0, obj=1.0, obj_pw=1.0, iou_t=0.2, anchor_t=4.0, fl_gamma=0.0, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, mosaic=1.0, mixup=0.0, copy_paste=0.0
Weights & Biases: run 'pip install wandb' to automatically track and visualize YOLOv5 🚀 runs (RECOMMENDED)
TensorBoard: Start with 'tensorboard --logdir runs/train', view at http://localhost:6006/
Overriding model.yaml nc=80 with nc=60

                 from  n    params  module                                  arguments                     
  0                -1  1      3520  models.common.Conv                      [3, 32, 6, 2, 2]              
  1                -1  1     18560  models.common.Conv                      [32, 64, 3, 2]                
  2                -1  1     18816  models.common.C3                        [64, 64, 1]                   
  3                -1  1     73984  models.common.Conv                      [64, 128, 3, 2]               
  4                -1  2    115712  models.common.C3                        [128, 128, 2]                 
  5                -1  1    295424  models.common.Conv                      [128, 256, 3, 2]              
  6                -1  3    625152  models.common.C3                        [256, 256, 3]                 
  7                -1  1   1180672  models.common.Conv                      [256, 512, 3, 2]              
  8                -1  1   1182720  models.common.C3                        [512, 512, 1]                 
  9                -1  1    656896  models.common.SPPF                      [512, 512, 5]                 
 10                -1  1    131584  models.common.Conv                      [512, 256, 1, 1]              
 11                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 12           [-1, 6]  1         0  models.common.Concat                    [1]                           
 13                -1  1    361984  models.common.C3                        [512, 256, 1, False]          
 14                -1  1     33024  models.common.Conv                      [256, 128, 1, 1]              
 15                -1  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 16           [-1, 4]  1         0  models.common.Concat                    [1]                           
 17                -1  1     90880  models.common.C3                        [256, 128, 1, False]          
 18                -1  1    147712  models.common.Conv                      [128, 128, 3, 2]              
 19          [-1, 14]  1         0  models.common.Concat                    [1]                           
 20                -1  1    296448  models.common.C3                        [256, 256, 1, False]          
 21                -1  1    590336  models.common.Conv                      [256, 256, 3, 2]              
 22          [-1, 10]  1         0  models.common.Concat                    [1]                           
 23                -1  1   1182720  models.common.C3                        [512, 512, 1, False]          
 24      [17, 20, 23]  1    175305  models.yolo.Detect                      [60, [[10, 13, 16, 30, 33, 23], [30, 61, 62, 45, 59, 119], [116, 90, 156, 198, 373, 326]], [128, 256, 512]]
Model Summary: 270 layers, 7181449 parameters, 7181449 gradients, 16.4 GFLOPs

Transferred 343/349 items from yolov5s.pt
Scaled weight_decay = 0.0005
optimizer: SGD with parameter groups 57 weight, 60 weight (no decay), 60 bias
train: Scanning '../datasets/xView/images/autosplit_train' images and labels...760 found, 0 missing, 0 empty, 0 corrupted: 100%|█| 760/760 [00:05<00:00, 139.20it/
train: New cache created: ../datasets/xView/images/autosplit_train.cache
val: Scanning '../datasets/xView/images/autosplit_val' images and labels...86 found, 0 missing, 0 empty, 0 corrupted: 100%|███████| 86/86 [00:06<00:00, 13.73it/s]
val: New cache created: ../datasets/xView/images/autosplit_val.cache
Plotting labels... 

etc...

[glenn-jocher]

@geobao also if I examine our indexing list I see 61 unique values includes -1, which means the indexing list is already incorporating 60 unique classes.

import numpy as np
np.unique(xview_class2index)
Out[6]: 
array([-1,  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
       16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32,
       33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49,
       50, 51, 52, 53, 54, 55, 56, 57, 58, 59])
len(np.unique(xview_class2index))
Out[7]: 61

[github-actions[bot]]

👋 Hello, this issue has been automatically marked as stale because it has not had recent activity. Please note it will be closed if no further activity occurs.

Access additional YOLOv5 🚀 resources:

• Wiki – https://github.com/ultralytics/yolov5/wiki
• Tutorials – https://docs.ultralytics.com/yolov5
• Docs – https://docs.ultralytics.com

Access additional Ultralytics ⚡ resources:

• Ultralytics HUB – https://ultralytics.com/hub
• Vision API – https://ultralytics.com/yolov5
• About Us – https://ultralytics.com/about
• Join Our Team – https://ultralytics.com/work
• Contact Us – https://ultralytics.com/contact

Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

Thank you for your contributions to YOLOv5 🚀 and Vision AI ⭐!

[hadign20]

@glenn-jocher How can I create the autosplit_train.txt and autosplit_val.txt files for xView? I cannot find them anywhere.

[glenn-jocher]

@hadi-ghnd you don't need to call autosplit(), this is done automatically when you first train on xView after downloading the data. Directions are in the yaml: https://github.com/ultralytics/yolov5/blob/19e0208fc9e33010717e066f9bd65c27db7c2b5c/data/xView.yaml#L1-L9

[hadign20]

@glenn-jocher ok, thank you. I downloaded the data and put it in a folder next to yolov5:

# ├── yolov5 
# └── datasets 
#     └── xView
            └── train_images
            └── train_labels
            └── val_images
            └── xView_train.geojson

Then I tried training with python train.py --data xView.yaml, but I got an error like this:

.... .... .... Converting ..\datasets\xView\xView_train.geojson: 98%|███████████████████████████████████████████████████████████████▊ | 590891/601937 [03:36<00:03, 3196.80it/s]WARNING: skipping one label for ..\datasets\xView\train_images\109.tif: incorrect class index -1 Converting ..\datasets\xView\xView_train.geojson: 98%|███████████████████████████████████████████████████████████████▊ | 591213/601937 [03:36<00:03, 3106.31it/s]WARNING: skipping one label for ..\datasets\xView\train_images\109.tif: incorrect class index -1 Converting ..\datasets\xView\xView_train.geojson: 100%|█████████████████████████████████████████████████████████████████| 601937/601937 [03:40<00:00, 2724.85it/s] Traceback (most recent call last): File "train.py", line 638, in main(opt) File "train.py", line 535, in main train(opt.hyp, opt, device, callbacks) File "train.py", line 95, in train loggers = Loggers(save_dir, weights, opt, hyp, LOGGER) # loggers instance File "D:\Project\online_packages\objectDetection\yolov5\utils\loggers__init.py", line 73, in init__ self.wandb = WandbLogger(self.opt, run_id) File "D:\Project\online_packages\objectDetection\yolov5\utils\loggers\wandb\wandb_utils.py", line 180, in init self.data_dict = check_wandb_dataset(opt.data) File "D:\Project\online_packages\objectDetection\yolov5\utils\loggers\wandb\wandb_utils.py", line 56, in check_wandb_dataset return check_dataset(data_file) File "D:\Project\online_packages\objectDetection\yolov5\utils\general.py", line 428, in check_dataset r = exec(s, {'yaml': data}) # return None File "", line 72, in File "D:\Project\online_packages\objectDetection\yolov5\utils\datasets.py", line 878, in autosplit [(path.parent / x).unlink(missing_ok=True) for x in txt] # remove existing File "D:\Project\online_packages\objectDetection\yolov5\utils\datasets.py", line 878, in [(path.parent / x).unlink(missing_ok=True) for x in txt] # remove existing TypeError: unlink() got an unexpected keyword argument 'missing_ok'

I thought maybe I should create the autosplit_val.txt file, but the error might be coming from wandb, or maybe my python version is low (3.7.12). I will try without wandb.

[glenn-jocher]

@hadi-ghnd your python version is likely too low.

[hadign20]

@glenn-jocher thank you. I updated the python version and disabled wandb. Now it's working, but the training is slow because of my GPU. Is there a way to get images like the one you posted? Or a pre-trained model on xView to test?

[glenn-jocher]

@hadi-ghnd there is no pretrained xView model.

Train batch images are generated automatically, i.e. run/train/exp/train_batch*.jpg

[hadign20]

@glenn-jocher thanks a lot. I see them.

[twangnh]

hi @glenn-jocher @hadi-ghnd thanks for your disscussion on the xView dataset issue, I encountered a similar problem, I have followed the instruction to download the dataset as :

datasets/
└── xView
    ├── train_images
    ├── train_images.zip
    ├── train_labels
    ├── train_labels.zip
    ├── val_images
    ├── val_images.zip
    └── xView_train.geojson

but when I run the train.py with python train.py --data xView.yaml --imgsz 640 --device 0, it raises:

train: weights=yolov5s.pt, cfg=, data=xView.yaml, hyp=data/hyps/hyp.scratch-low.yaml, epochs=300, batch_size=16, imgsz=640, rect=False, resume=False, nosave=False, noval=False, noautoanchor=True, noplots=False, evolve=None, bucket=, cache=None, image_weights=False, device=0, multi_scale=False, single_cls=False, optimizer=SGD, sync_bn=False, workers=8, project=runs/train, name=exp, exist_ok=False, quad=False, cos_lr=False, label_smoothing=0.0, patience=100, freeze=[0], save_period=-1, seed=0, local_rank=-1, entity=None, upload_dataset=False, bbox_interval=-1, artifact_alias=latest
default message here: github: skipping check (offline), for updates see https://github.com/ultralytics/yolov5
YOLOv5 🚀 v6.2-73-g91a81d4 Python-3.8.3 torch-1.8.1+cu101 CUDA:0 (GeForce GTX TITAN X, 12212MiB)

hyperparameters: lr0=0.01, lrf=0.01, momentum=0.937, weight_decay=0.0005, warmup_epochs=3.0, warmup_momentum=0.8, warmup_bias_lr=0.1, box=0.05, cls=0.5, cls_pw=1.0, obj=1.0, obj_pw=1.0, iou_t=0.2, anchor_t=4.0, fl_gamma=0.0, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, degrees=0.0, translate=0.1, scale=0.5, shear=0.0, perspective=0.0, flipud=0.0, fliplr=0.5, mosaic=1.0, mixup=0.0, copy_paste=0.0
Weights & Biases: run 'pip install wandb' to automatically track and visualize YOLOv5 🚀 runs in Weights & Biases
ClearML: run 'pip install clearml' to automatically track, visualize and remotely train YOLOv5 🚀 in ClearML
TensorBoard: Start with 'tensorboard --logdir runs/train', view at http://localhost:6006/

Dataset not found ⚠️, missing paths ['/home/wangtao/prj/datasets/xView/images/autosplit_val.txt']
Traceback (most recent call last):
  File "/home/wangtao/anaconda3_2/envs/deform_seg_env/lib/python3.8/contextlib.py", line 131, in __exit__
    self.gen.throw(type, value, traceback)
  File "/home/wangtao/prj/yolov5/utils/torch_utils.py", line 92, in torch_distributed_zero_first
    yield
  File "/home/wangtao/prj/yolov5/train.py", line 122, in train
    data_dict = data_dict or check_dataset(data)  # check if None
  File "/home/wangtao/prj/yolov5/utils/general.py", line 516, in check_dataset
    r = exec(s, {'yaml': data})  # return None
  File "<string>", line 9, in <module>
ModuleNotFoundError: No module named 'utils.datasets'

[glenn-jocher]

@twangnh good news 😃! Your original issue may now be fixed ✅ in PR #9807.

To receive this update:

• Git – git pull from within your yolov5/ directory or git clone https://github.com/ultralytics/yolov5 again
• PyTorch Hub – Force-reload model = torch.hub.load('ultralytics/yolov5', 'yolov5s', force_reload=True)
• Notebooks – View updated notebooks
• Docker – sudo docker pull ultralytics/yolov5:latest to update your image

Thank you for spotting this issue and informing us of the problem. Please let us know if this update resolves the issue for you, and feel free to inform us of any other issues you discover or feature requests that come to mind. Happy trainings with YOLOv5 🚀!