Skip to content

By the end of this post, you will learn how to: Train a SOTA YOLOv5 model on your own data. Sparsify the model using SparseML quantization aware training, sparse transfer learning, and one-shot quantization. Export the sparsified model and run it using the DeepSparse engine at insane speeds. P/S: The end result - YOLOv5 on CPU at 180+ FPS using on

Notifications You must be signed in to change notification settings

dnth/yolov5-deepsparse-blogpost

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

74 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Supercharging YOLOv5: How I Got 182.4 FPS Inference Without a GPU

image

Companion repo for the blogpost.

❤️ Support Me

If you like what you see, support me in keeping the lights on to produce more posts like this.

Buy Me A Coffee

🔗 Installation

git clone https://github.com/dnth/yolov5-deepsparse-blogpost
cd yolov5-deepsparse-blogpost/
pip install torch==1.9.0 torchvision==0.10.0 --extra-index-url https://download.pytorch.org/whl/cu111
pip install -r req.txt

Or (Highly recommended!) 👇

🔥 Run In Colab

The easiest way to get started is to run this Colab Notebook.

The notebook serves as a guide to:

  • Install all packages used this blog post.
  • Train a sparse YOLOv5 models using SparseML.
  • Run inference using the DeepSparse engine.

🥋 Training

YOLOv5-S Baseline

python train.py --cfg ./models_v5.0/yolov5s.yaml --data pistols.yaml --hyp data/hyps/hyp.scratch.yaml --weights yolov5s.pt --img 416 --batch-size 64 --optimizer SGD --epochs 100 --device 0 --project yolov5-deepsparse --name yolov5s-sgd

YOLOv5-S (One-Shot)

python train.py --cfg ./models_v5.0/yolov5s.yaml --recipe ../recipes/yolov5s.pruned.md --data pistols.yaml --hyp data/hyps/hyp.scratch.yaml --weights yolov5-deepsparse/yolov5s-sgd/weights/best.pt --img 416 --batch-size 64 --optimizer SGD --epochs 100 --device 0 --project yolov5-deepsparse --name yolov5s-sgd-one-shot --one-shot

YOLOv5-S Pruned

python train.py --cfg ./models_v5.0/yolov5s.yaml --recipe ../recipes/yolov5s.pruned.md --data pistols.yaml --hyp data/hyps/hyp.scratch.yaml --weights yolov5s.pt --img 416 --batch-size 64 --optimizer SGD --device 0 --project yolov5-deepsparse --name yolov5s-sgd-pruned

YOLOv5-S Quantized

python train.py --cfg ./models_v5.0/yolov5s.yaml --recipe ../recipes/yolov5s.quantized.md --data pistols.yaml --hyp data/hyps/hyp.scratch.yaml --weights yolov5-deepsparse/yolov5s-sgd/weights/best.pt --img 416 --batch-size 64 --project yolov5-deepsparse --name yolov5s-sgd-quantized

YOLOv5-S Pruned + Quantized

python train.py --cfg ./models_v5.0/yolov5s.yaml --recipe ../recipes/yolov5.transfer_learn_pruned_quantized.md --data pistols.yaml --hyp data/hyps/hyp.scratch.yaml --weights yolov5s.pt --img 416 --batch-size 64 --optimizer SGD --device 0 --project yolov5-deepsparse --name yolov5s-sgd-pruned-quantized

YOLOv5-S Transfer Learning

python train.py --data pistols.yaml --cfg ./models_v5.0/yolov5s.yaml --weights zoo:cv/detection/yolov5-s/pytorch/ultralytics/coco/pruned_quant-aggressive_94?recipe_type=transfer --img 416 --batch-size 64 --hyp data/hyps/hyp.scratch.yaml --recipe ../recipes/yolov5.transfer_learn_pruned_quantized.md --optimizer SGD --device 0 --project yolov5-deepsparse --name yolov5s-sgd-pruned-quantized-transfer

YOLOv5n Pruned + Quantized

python train.py --cfg ./models_v5.0/yolov5n.yaml --recipe ../recipes/yolov5.transfer_learn_pruned_quantized.md --data pistols.yaml --hyp data/hyps/hyp.scratch.yaml --weights yolov5n.pt --img 416 --batch-size 64 --optimizer SGD --device 0 --project yolov5-deepsparse --name yolov5n-sgd-pruned-quantized

🤖 Export to ONNX

YOLOv5-S Baseline

python export.py --weights yolov5-deepsparse/yolov5s-sgd/weights/best.pt --include onnx --imgsz 416 --dynamic --simplify

YOLOv5-S (One-Shot)

python export.py --weights yolov5-deepsparse/yolov5s-sgd-one-shot/weights/checkpoint-one-shot.pt --include onnx --imgsz 416 --dynamic --simplify

YOLOv5-S Pruned

python export.py --weights yolov5-deepsparse/yolov5s-sgd-pruned/weights/best.pt --include onnx --imgsz 416 --dynamic --simplify

YOLOv5-S Quantized

python export.py --weights yolov5-deepsparse/yolov5s-sgd-quantized/weights/best.pt --include onnx --imgsz 416 --dynamic --simplify

YOLOv5-S Pruned + Quantized

python export.py --weights yolov5-deepsparse/yolov5s-sgd-pruned-quantized/weights/best.pt --include onnx --imgsz 416 --dynamic --simplify

YOLOv5-S Transfer Learning

python export.py --weights yolov5-deepsparse/yolov5s-sgd-pruned-quantized-transfer/weights/best.pt --include onnx --imgsz 416 --dynamic --simplify

YOLOv5n Pruned + Quantized

python export.py --weights yolov5-deepsparse/yolov5n-sgd-pruned-quantized/weights/best.pt --include onnx --imgsz 416 --dynamic --simplify

🚀 Inference

YOLOv5-S Baseline - PyTorch Engine

python annotate.py yolov5-deepsparse/yolov5s-sgd/weights/best.pt --source data/pexels-cottonbro-8717592.mp4 --engine torch --image-shape 416 416 --device cpu --conf-thres 0.7

YOLOv5-S Baseline - DeepSparse Engine

python annotate.py yolov5-deepsparse/yolov5s-sgd/weights/best.onnx --source data/pexels-cottonbro-8717592.mp4 --engine deepsparse --device cpu --conf-thres 0.7 --image-shape 416 416 --num-cores 4

YOLOv5-S (One-Shot) - DeepSparse Engine

python annotate.py yolov5-deepsparse/yolov5s-sgd-one-shot/weights/checkpoint-one-shot.onnx --source data/pexels-cottonbro-8717592.mp4 --engine deepsparse --device cpu --conf-thres 0.7 --image-shape 416 416 --num-cores 4

YOLOv5-S Pruned - DeepSparse Engine

python annotate.py yolov5-deepsparse/yolov5s-sgd-pruned/weights/best.onnx --source data/pexels-cottonbro-8717592.mp4 --engine deepsparse --device cpu --conf-thres 0.7 --image-shape 416 416 --num-cores 4

YOLOv5-S Quantized - DeepSparse Engine

python annotate.py yolov5-deepsparse/yolov5s-sgd-quantized/weights/best.onnx --source data/pexels-cottonbro-8717592.mp4 --engine deepsparse --device cpu --conf-thres 0.7 --image-shape 416 416 --quantized-input --num-cores 4

YOLOv5-S Pruned + Quantized - DeepSparse Engine

python annotate.py yolov5-deepsparse/yolov5s-sgd-pruned-quantized/weights/best.onnx --source data/pexels-cottonbro-8717592.mp4 --engine deepsparse --device cpu --conf-thres 0.7 --image-shape 416 416 --quantized-input --num-cores 4

YOLOv5-S Transfer Learning - DeepSparse Engine

python annotate.py yolov5-deepsparse/yolov5s-sgd-pruned-quantized-transfer/weights/best.onnx --source data/pexels-cottonbro-8717592.mp4 --engine deepsparse --device cpu --conf-thres 0.8 --image-shape 416 416 --quantized-input --num-cores 4

YOLOv5-Nano Pruned + Quantized - DeepSparse Engine

python annotate.py yolov5-deepsparse/yolov5n-sgd-pruned-quantized-hardswish/weights/best.onnx --source data/pexels-cottonbro-8717592.mp4 --engine deepsparse --device cpu --conf-thres 0.7 --image-shape 416 416 --quantized-input --num-cores 4

Wandb Dashboard

https://wandb.ai/dnth/yolov5-deepsparse?workspace=user-dnth

Detect

python detect.py --weights yolov5-deepsparse/yolov5s-sgd/weights/best.pt --source data/pexels-cottonbro-8717592.mp4 --data data/pistols.yaml --imgsz 416 --view-img --nosave --device cpu

About

By the end of this post, you will learn how to: Train a SOTA YOLOv5 model on your own data. Sparsify the model using SparseML quantization aware training, sparse transfer learning, and one-shot quantization. Export the sparsified model and run it using the DeepSparse engine at insane speeds. P/S: The end result - YOLOv5 on CPU at 180+ FPS using on

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published