Contents

• Contents
• Darknet53 Description
  • Description
  • Paper
• Model Architecture
• Dataset
• Features
  • Mixed Precision
• Environment Requirements
• Quick Start
• Script Description
  • Script and Sample Code
  • Script Parameters
  • Training Process
    • Usage
      • Running on GPU
  • Evaluation Process
    • Usage
      • Running on GPU
    • Result
• Model Description
  • Performance
    • Evaluation Performance
      • Darknet53 on ImageNet2012
    • Inference Performance
      • Darknet53 on ImageNet2012
• Description of Random Situation
• ModelZoo Homepage

Darknet53 Description

Description

"Darknet53" is a backbone for Yolov3 model.

Paper

paper:Joseph Redmon, Ali Farhadi. "YOLOv3: An Incremental Improvement"

Model Architecture

The overall network architecture of Net is show below: Link

Dataset

Dataset used: ImageNet2012

• Dataset size 224*224 colorful images in 1000 classes
  • Train：1,281,167 images
  • Test： 50,000 images
• Data format：jpeg
  • Note：Data will be processed in dataset.py

Features

Mixed Precision

The mixed precision training method accelerates the deep learning neural network training process by using both the single-precision and half-precision data types, and maintains the network precision achieved by the single-precision training at the same time. Mixed precision training can accelerate the computation process, reduce memory usage, and enable a larger model or batch size to be trained on specific hardware. For FP16 operators, if the input data type is FP32, the backend of MindSpore will automatically handle it with reduced precision. Users could check the reduced-precision operators by enabling INFO log and then searching ‘reduce precision’.

Environment Requirements

• Hardware（GPU）
  • Prepare hardware environment GPU processor.
• Framework
  • MindSpore
• For more information, please check the resources below：
  • MindSpore Tutorials
  • MindSpore Python API

Quick Start

After installing MindSpore via the official website, you can start training and evaluation as follows:

• Running on GPU

# distributed training
Usage:
bash run_distribute_train_gpu.sh darknet53 imagenet2012 [DATASET_PATH]

# standalone training
Usage:
bash run_standalone_train_gpu.sh darknet53 imagenet2012 [DATASET_PATH]

# run evaluation example
Usage:
bash run_eval_gpu.sh [DATASET_PATH] [CHECKPOINT_PATH]

Script Description

Script and Sample Code

.
└──darknet53
  ├── README.md
  ├── scripts
    ├── run_distribute_train_gpu.sh        # launch gpu distributed training(8 pcs)
    ├── run_eval_gpu.sh                    # launch gpu evaluation
    └── run_standalone_train_gpu.sh        # launch gpu standalone training(1 pcs)
  ├── src
    ├── config.py                          # parameter configuration
    ├── CrossEntropySmooth.py              # loss definition for ImageNet2012 dataset
    ├── dataset.py                         # data preprocessing
    ├── lr_generator.py                    # generate learning rate for each step
    ├── callbacks.py                       # callbacks for training
    └── darknet53.py                       # darknet53
  ├── export.py                            # export model for inference
  ├── eval.py                              # eval net
  └── train.py                             # train net

Script Parameters

Parameters for both training and evaluation can be set in imagenet_config.yaml

• Config for Darknet53, ImageNet2012 dataset

device_target: "GPU"
run_distributed: True
device_num: ""
net: "darknet53"
dataset: "imagenet2012"
dataset_path: ""
pre_trained: ""
class_num: 1000
epoch_num: 150
pretrained_epoch_num: 0
batch_size: 256
momentum: 0.9
loss_scale: 1024
weight_decay: 0.0001
use_label_smooth: True
label_smooth_factor: 0.1
lr: 0.1
lr_scheduler: "cosine_annealing"
eta_min: 0
T_max: 150
warmup_epochs: 5
checkpoint_path: ""
save_checkpoint: True
save_checkpoint_epochs: 5
keep_checkpoint_max: 90
save_checkpoint_path: "./"
summary_dir: "./summary_dir"

Training Process

Usage

Mindinsight support

You can open http://127.0.0.1:8080 to view trainig process (training loss and validation accuracy). More info you can find in the link MindInsight

Running on GPU

# distributed training
Usage:
bash run_distribute_train_gpu.sh darknet53 imagenet2012 [DATASET_PATH]

# standalone training
Usage:
bash run_standalone_train_gpu.sh darknet53 imagenet2012 [DATASET_PATH]

Training result will be stored in the example path, whose folder name begins with "train" or "train_parallel". Under this, you can find checkpoint file together with result like the following in log.

Evaluation Process

Usage

Running on GPU

bash run_eval_gpu.sh [DATASET_PATH] [CHECKPOINT_PATH]

Result

• Evaluating Darknet53 with ImageNet2012 dataset

result: {'top_5_accuracy': 0.9247195512820513, 'top_1_accuracy': 0.7572916666666667}

Inference Process

Export MindIR

python export.py --ckpt_file [CKPT_PATH] --batch_size [BATCH_SIZE] --file_format [FILE_FORMAT]

The ckpt_file parameter is required, FILE_FORMAT should be in ["AIR", "MINDIR"] BATCH_SIZE current batch_size can only be set to 1.

Model Description

Performance

Evaluation Performance

Darknet53 on ImageNet2012

Parameter	Darknet53
Resource	GPU: 8xRTX3090 24G; CPU: Intel(R) Xeon(R) Gold 6226R; RAM: 252G
Upload date	2021-11-03
MindSpore version	1.3.0
Dataset	ImageNet
Training parameters	imagenet_config.yaml
Optimizer	Momentum
Loss function	SoftmaxCrossEntropy
Output	ckpt file
Final loss	1.33
Top1 Accuracy	75.7%
Total time	135h
Parameters (M)	batch_size=256, epoch=150

Inference Performance

Darknet53 on ImageNet2012

Parameter	Darknet53
Resource	GPU: 8xRTX3090 24G; CPU: Intel(R) Xeon(R) Gold 6226R; RAM: 252G
Upload date	2021-11-03
MindSpore version	1.3.0
Dataset	ImageNet
Batch size	256（1card）
Output	Probability
Accuracy	ACC1[75.7%] ACC5[92.5%]
Total time	4m

Description of Random Situation

In dataset.py, we set the seed inside "create_dataset" function. We also use random seed in train.py.

ModelZoo Homepage

Please check the official homepage.