Ankle/yolov5_ZWgra
1
0 Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
Files
e16e9e43e1b4b8f9b9819fe81d537c00ec6d606e
yolov5_ZWgra/tutorial.ipynb
T

3.1 MiB
Raw Blame History

Open In Colab

This notebook was developed by Ultralytics LLC, and is freely available for redistribution under the GPL-3.0 license. For more information please visit https://github.com/ultralytics/yolov5 and https://www.ultralytics.com.

#Initial Setup

Clone repo, install dependencies, %cd into ./yolov5 folder and check GPU.

In [0]:
!git clone https://github.com/ultralytics/yolov5  # clone repo
!pip install -r yolov5/requirements.txt  # install dependencies
%cd yolov5

import torch
from IPython.display import Image, clear_output  # to display images
from utils.google_utils import gdrive_download  # to download models/datasets

clear_output()
print('Setup complete. Using torch %s %s' % (torch.__version__, torch.cuda.get_device_properties(0) if torch.cuda.is_available() else 'CPU'))

#1. Inference

Run inference with a pretrained checkpoint on contents of /inference/images folder. Models are downloaded automatically from our Google Drive folder if available.

In [14]:
!python detect.py --weights yolov5s.pt --img 416 --conf 0.4 --source ./inference/images/
Image(filename='inference/output/zidane.jpg', width=600)
Out [14]:
Namespace(agnostic_nms=False, augment=False, classes=None, conf_thres=0.4, device='', fourcc='mp4v', half=False, img_size=416, iou_thres=0.5, output='inference/output', save_txt=False, source='./inference/images/', view_img=False, weights='yolov5s.pt')
Using CUDA device0 _CudaDeviceProperties(name='Tesla P100-PCIE-16GB', total_memory=16280MB)

image 1/2 inference/images/bus.jpg: 416x352 3 persons, 1 buss, Done. (0.009s)
image 2/2 inference/images/zidane.jpg: 288x416 2 persons, 2 ties, Done. (0.009s)
Results saved to /content/yolov5/inference/output
Done. (0.100s)

Inference can be run on a variety of sources: images, videos, directories, webcams, rtsp and http streams as shown in the example below.

In [0]:
# Example syntax (do not run cell)
!python detect.py --source ./file.jpg  # image 
                           ./file.mp4  # video
                           ./dir  # directory
                           0  # webcam
                           'rtsp://170.93.143.139/rtplive/470011e600ef003a004ee33696235daa' # rtsp
                           'http://112.50.243.8/PLTV/88888888/224/3221225900/1.m3u8'  # http

#2. Test Test a model on COCO val or test-dev dataset to determine trained accuracy. Models are downloaded automatically from our Google Drive folder if available. To show results by class use the --verbose flag. Note that pycocotools metrics may be 1-2% better than the equivalent repo metrics, as is visible below, due to slight differences in mAP computation.

###2.1 val2017 Download COCO val 2017 dataset, 1GB, 5000 images, and test model accuracy.

In [7]:
# Download COCO val2017
gdrive_download('1Y6Kou6kEB0ZEMCCpJSKStCor4KAReE43','coco2017val.zip')  # val2017 dataset
!mv ./coco ../  # move folder alongside /yolov5
Downloading https://drive.google.com/uc?export=download&id=1Y6Kou6kEB0ZEMCCpJSKStCor4KAReE43 as coco2017val.zip... unzipping... Done (37.1s)
In [8]:
# Run YOLOv5s on COCO val2017
!python test.py --weights yolov5s.pt --data ./data/coco.yaml --img 640
Namespace(augment=False, batch_size=32, conf_thres=0.001, data='././data/coco.yaml', device='', img_size=640, iou_thres=0.65, save_json=True, single_cls=False, task='val', verbose=False, weights='yolov5s.pt')
Using CUDA device0 _CudaDeviceProperties(name='Tesla T4', total_memory=15079MB)

Model Summary: 165 layers, 7.07417e+06 parameters, 7.07417e+06 gradients
Caching labels ../coco/val2017.txt (4952 found, 48 missing, 0 empty, 0 duplicate, for 5000 images): 100% 5000/5000 [00:00<00:00, 7822.09it/s]
Saving labels to ../coco/labels/val2017.npy for faster future loading
               Class      Images     Targets           P           R      mAP@.5  mAP@.5:.95: 100% 157/157 [01:33<00:00,  1.67it/s]
                 all       5e+03    3.63e+04       0.334       0.633       0.534       0.336
Speed: 6.0/2.9/8.9 ms inference/NMS/total per 640x640 image at batch-size 32

COCO mAP with pycocotools... saving detections_val2017_yolov5s_results.json...
loading annotations into memory...
Done (t=0.41s)
creating index...
index created!
Loading and preparing results...
DONE (t=5.70s)
creating index...
index created!
Running per image evaluation...
Evaluate annotation type *bbox*
DONE (t=89.04s).
Accumulating evaluation results...
DONE (t=14.04s).
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.352
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.544
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.378
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.188
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.397
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.459
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.296
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.496
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.557
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.358
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.618
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.700
In [10]:
# Run YOLOv5x on COCO val2017
!python test.py --weights yolov5x.pt --data ./data/coco.yaml --img 640
Namespace(augment=False, batch_size=32, conf_thres=0.001, data='././data/coco.yaml', device='', img_size=640, iou_thres=0.65, save_json=True, single_cls=False, task='val', verbose=False, weights='yolov5x.pt')
Using CUDA device0 _CudaDeviceProperties(name='Tesla T4', total_memory=15079MB)

Model Summary: 381 layers, 9.59219e+07 parameters, 9.59219e+07 gradients
Caching labels ../coco/labels/val2017.npy (4952 found, 0 missing, 48 empty, 0 duplicate, for 5000 images): 100% 5000/5000 [00:00<00:00, 16516.13it/s]
               Class      Images     Targets           P           R      mAP@.5  mAP@.5:.95: 100% 157/157 [05:05<00:00,  1.95s/it]
                 all       5e+03    3.63e+04       0.393       0.742       0.652       0.455
Speed: 51.1/2.2/53.3 ms inference/NMS/total per 640x640 image at batch-size 32

COCO mAP with pycocotools... saving detections_val2017_yolov5x_results.json...
loading annotations into memory...
Done (t=0.39s)
creating index...
index created!
Loading and preparing results...
DONE (t=3.59s)
creating index...
index created!
Running per image evaluation...
Evaluate annotation type *bbox*
DONE (t=79.47s).
Accumulating evaluation results...
DONE (t=12.20s).
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.470
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.659
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.515
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.310
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.516
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.610
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.362
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.597
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.656
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.504
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.704
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.787

###2.2 test-dev2017 Download COCO test2017 dataset, 7GB, 40,000 images, to test model accuracy on test-dev set, 20,000 images. Results are saved to a *.json file which can be submitted to the evaluation server at https://competitions.codalab.org/competitions/20794.

In [0]:
# Download COCO test-dev2017
gdrive_download('1cXZR_ckHki6nddOmcysCuuJFM--T-Q6L','coco2017labels.zip')  # annotations
!f="test2017.zip" && curl http://images.cocodataset.org/zips/$f -o $f && unzip -q $f && rm $f  # 7GB,  41k images
!mv ./test2017 ./coco/images && mv ./coco ../  # move images into /coco and move /coco alongside /yolov5
In [0]:
# Run YOLOv5s on COCO test-dev2017 with argument --task test
!python test.py --weights yolov5s.pt --data ./data/coco.yaml --task test

3. Train

Download the 128-image tutorial training dataset ./data/coco128.yaml, start tensorboard and train a yolov5s.yaml model for 5 epochs. Note that actual training is typically much longer, around 300-1000 epochs, depending on your dataset.

In [2]:
# Download tutorial dataset coco128.yaml
gdrive_download('1n_oKgR81BJtqk75b00eAjdv03qVCQn2f','coco128.zip')  # tutorial dataset
!mv ./coco128 ../  # move folder alongside /yolov5
Downloading https://drive.google.com/uc?export=download&id=1n_oKgR81BJtqk75b00eAjdv03qVCQn2f as coco128.zip... unzipping... Done (5.8s)
In [0]:
# Start tensorboard
%load_ext tensorboard
%tensorboard --logdir runs

Train a YOLOv5s model on the coco128 dataset by specifying model configuration file --cfg ./models/yolo5s.yaml, and a dataset configuration file --data ./data/coco128.yaml. Start training from pretrained --weights yolov5s.pt, or from scratch (randomly initialized weights) using --weights ''. Pretrained checkpoints are downloaded automatically from our Google Drive folder if available.

In [5]:
# Train YOLOv5s on coco128 for 5 epochs
!python train.py --img 640 --batch 16 --epochs 5 --data ./data/coco128.yaml --cfg ./models/yolov5s.yaml --weights yolov5s.pt --name tutorial --nosave --cache
Apex recommended for faster mixed precision training: https://github.com/NVIDIA/apex
{'lr0': 0.01, 'momentum': 0.937, 'weight_decay': 0.0005, 'giou': 0.05, 'cls': 0.58, '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': 0.014, 'hsv_s': 0.68, 'hsv_v': 0.36, 'degrees': 0.0, 'translate': 0.0, 'scale': 0.5, 'shear': 0.0}
Namespace(adam=False, batch_size=16, bucket='', cache_images=True, cfg='././models/yolov5s.yaml', data='././data/coco128.yaml', device='', epochs=5, evolve=False, img_size=[640], multi_scale=False, name='tutorial', nosave=True, notest=False, rect=False, resume=False, single_cls=False, weights='yolov5s.pt')
Using CUDA device0 _CudaDeviceProperties(name='Tesla T4', total_memory=15079MB)

2020-06-14 19:45:19.668043: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.1
Start Tensorboard with "tensorboard --logdir=runs", view at http://localhost:6006/

              from  n    params  module                                  arguments                     
  0             -1  1      3520  models.common.Focus                     [3, 32, 3]                    
  1             -1  1     18560  models.common.Conv                      [32, 64, 3, 2]                
  2             -1  1     20672  models.common.Bottleneck                [64, 64]                      
  3             -1  1     73984  models.common.Conv                      [64, 128, 3, 2]               
  4             -1  1    161152  models.common.BottleneckCSP             [128, 128, 3]                 
  5             -1  1    295424  models.common.Conv                      [128, 256, 3, 2]              
  6             -1  1    641792  models.common.BottleneckCSP             [256, 256, 3]                 
  7             -1  1   1180672  models.common.Conv                      [256, 512, 3, 2]              
  8             -1  1    656896  models.common.SPP                       [512, 512, [5, 9, 13]]        
  9             -1  1   1905152  models.common.BottleneckCSP             [512, 512, 2]                 
 10             -1  1   1248768  models.common.BottleneckCSP             [512, 512, 1, False]          
 11             -1  1    130815  torch.nn.modules.conv.Conv2d            [512, 255, 1, 1]              
 12             -2  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 13        [-1, 6]  1         0  models.common.Concat                    [1]                           
 14             -1  1    197120  models.common.Conv                      [768, 256, 1, 1]              
 15             -1  1    313088  models.common.BottleneckCSP             [256, 256, 1, False]          
 16             -1  1     65535  torch.nn.modules.conv.Conv2d            [256, 255, 1, 1]              
 17             -2  1         0  torch.nn.modules.upsampling.Upsample    [None, 2, 'nearest']          
 18        [-1, 4]  1         0  models.common.Concat                    [1]                           
 19             -1  1     49408  models.common.Conv                      [384, 128, 1, 1]              
 20             -1  1     78720  models.common.BottleneckCSP             [128, 128, 1, False]          
 21             -1  1     32895  torch.nn.modules.conv.Conv2d            [128, 255, 1, 1]              
 22   [-1, 16, 11]  1         0  models.yolo.Detect                      [80, [[10, 13, 16, 30, 33, 23], [30, 61, 62, 45, 59, 119], [116, 90, 156, 198, 373, 326]]]
Model Summary: 165 layers, 7.07417e+06 parameters, 7.07417e+06 gradients

Optimizer groups: 54 .bias, 60 conv.weight, 51 other
Reading image shapes: 100% 128/128 [00:00<00:00, 8158.02it/s]
Caching labels ../coco128/labels/train2017 (126 found, 0 missing, 2 empty, 0 duplicate, for 128 images): 100% 128/128 [00:00<00:00, 7608.18it/s]
Caching images (0.1GB): 100% 128/128 [00:00<00:00, 149.70it/s]
Caching labels ../coco128/labels/train2017 (126 found, 0 missing, 2 empty, 0 duplicate, for 128 images): 100% 128/128 [00:00<00:00, 8165.09it/s]
Caching images (0.1GB): 100% 128/128 [00:00<00:00, 145.34it/s]
Label width-height:         n      mean       min       max  matching    recall
                          929     105.8      1.23       640    0.4736    0.9903
Image sizes 640 train, 640 test
Using 2 dataloader workers
Starting training for 5 epochs...

     Epoch   gpu_mem      GIoU       obj       cls     total   targets  img_size
       0/4      7.3G   0.04362   0.07954   0.02032    0.1435       208       640: 100% 8/8 [00:09<00:00,  1.24s/it]
               Class      Images     Targets           P           R      mAP@.5  mAP@.5:.95: 100% 8/8 [00:07<00:00,  1.03it/s]
                 all         128         929       0.421       0.704        0.66        0.42

     Epoch   gpu_mem      GIoU       obj       cls     total   targets  img_size
       1/4     13.9G   0.04609   0.07819   0.01886    0.1431       167       640: 100% 8/8 [00:03<00:00,  2.34it/s]
               Class      Images     Targets           P           R      mAP@.5  mAP@.5:.95: 100% 8/8 [00:02<00:00,  3.34it/s]
                 all         128         929       0.418       0.713       0.659       0.421

     Epoch   gpu_mem      GIoU       obj       cls     total   targets  img_size
       2/4     13.9G   0.04508   0.07012   0.01977     0.135       198       640: 100% 8/8 [00:03<00:00,  2.33it/s]
               Class      Images     Targets           P           R      mAP@.5  mAP@.5:.95: 100% 8/8 [00:02<00:00,  3.50it/s]
                 all         128         929       0.409        0.71       0.664       0.421

     Epoch   gpu_mem      GIoU       obj       cls     total   targets  img_size
       3/4     13.9G   0.04577   0.08308   0.02038    0.1492       222       640: 100% 8/8 [00:03<00:00,  2.34it/s]
               Class      Images     Targets           P           R      mAP@.5  mAP@.5:.95: 100% 8/8 [00:02<00:00,  3.50it/s]
                 all         128         929       0.429        0.71        0.67       0.426

     Epoch   gpu_mem      GIoU       obj       cls     total   targets  img_size
       4/4     13.9G   0.04608   0.07352   0.01855    0.1382       149       640: 100% 8/8 [00:03<00:00,  2.32it/s]
               Class      Images     Targets           P           R      mAP@.5  mAP@.5:.95: 100% 8/8 [00:02<00:00,  3.46it/s]
                 all         128         929       0.423       0.716       0.663       0.422
Optimizer stripped from weights/last_tutorial.pt
5 epochs completed in 0.012 hours.

#4. Visualize

After training starts, view train*.jpg images to see training images, labels and augmentation effects. Note a mosaic dataloader is used for training (shown below), a new dataloading concept developed by Ultralytics and first featured in YOLOv4.

In [0]:
Image(filename='./train_batch1.jpg', width=900)  # view augmented training mosaics
Out [0]:

View test_batch0_gt.jpg to see test batch 0 ground truth labels.

In [0]:
Image(filename='./test_batch0_gt.jpg', width=900)  # view test image labels
Out [0]:

View test_batch0_pred.jpg to see test batch 0 predictions.

In [0]:
Image(filename='./test_batch0_pred.jpg', width=900)  # view test image predictions
Out [0]:

Training losses and performance metrics are saved to Tensorboard and also to a results.txt logfile. results.txt is plotted as results.png after training completes. Partially completed results.txt files can be plotted with from utils.utils import plot_results; plot_results(). Here we show YOLOv5s trained on coco128 to 100 epochs, starting from scratch (orange), and starting from pretrained yolov5s.pt weights (blue).

In [0]:
from utils.utils import plot_results; plot_results()  # plot results.txt as results.png
Image(filename='./results.png', width=1000)  # view results.png
Out [0]:

5. Appendix

In [0]:
# Re-clone
%cd ..
!rm -rf yolov5 && git clone https://github.com/ultralytics/yolov5
%cd yolov5
In [0]:
# Apex install
git clone https://github.com/NVIDIA/apex && cd apex && pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" . --user && cd .. && rm -rf apex
In [0]:
# Test GCP checkpoint on COCO val2017
%%shell
x=best*.pt
gsutil cp gs://*/*/weights/$x .
python test.py --weights $x --data ./data/coco.yaml --img 736
In [0]:
# Test multiple models on COCO val2017
%%shell
for x in yolov5s yolov5m yolov5l yolov5x
do 
  python test.py --weights $x.pt --data ./data/coco.yaml --img 640 --conf 0.001
done
Reference in new issue View Git Blame Copy Permalink