基于PaddleClas的Swin Transformer多标签图像分类

AI Studio

318人浏览 · 2023-05-05 11:14:37

AI Studio · 2023-05-05 11:14:37 发布

★★★ 本文源自AlStudio社区精品项目，【点击此处】查看更多精品内容 >>>

1.情况简介

该项目基于PaddleClas，主要完成多标签分类的训练、评估、预测的体验过程。PaddleClas的多标签分类模型只有MobileNetV1，因此本项目是基于MobileNetV1来进行改写SwinTransformer模型。

2.数据集

该项目数据集为NUS-WIDE-SCENE的子集，需要对图像进行分类，具有36个标签。

该子集下载地址： https://paddle-imagenet-models-name.bj.bcebos.com/data/NUS-SCENE-dataset.tar
NUS-WIDE-SCENE数据集下载地址：https://lms.comp.nus.edu.sg/wp-content/uploads/2019/research/nuswide/NUS-WIDE.html

标签有：

airport
beach
bridge
buildings
castle
cityscape
clouds
frost

garden
glacier
grass
harbor
house
lake
moon
mountain
nighttime
ocean
plants
railroad
rainbow
reflection
road
sky
snow
street
sunset
temple
town
valley
water
waterfall
window

二、PaddleClas安装

1.PaddleClas下载

#可以直接从套件中下载PaddleClas,这样速度比较快，这里下载v2.4.0版本。

2.PaddleClas安装

主要完成相关依赖库安装等

!pip install -r ~/PaddleClas-2.4.0/requirements.txt >log.log
!pip install -e ~/PaddleClas-2.4.0 >log.log

三、数据集准备

1.数据解压缩

主要完成 数据集下载、解压缩 等。

%cd ~/PaddleClas-2.4.0
!mkdir dataset/NUS-WIDE-SCENE
%cd dataset/NUS-WIDE-SCENE
!wget https://paddle-imagenet-models-name.bj.bcebos.com/data/NUS-SCENE-dataset.tar
!tar -xf NUS-SCENE-dataset.tar

2.数据查看

其中第一列为图像文件名，其后36列分别为garden glacier grass harbor house lake moon mountain nighttime ocean plants railroad rainbow reflection road sky snow street sunset temple town valley water waterfall window 标签，为1则是，0否。

!head NUS-SCENE-dataset/multilabel_train_list.txt

from PIL import Image
%cd ~
img=Image.open("PaddleClas-2.4.0/dataset/NUS-WIDE-SCENE/NUS-SCENE-dataset/images/0006_2074187535.jpg")
img.show()

四、模型训练

1.训练配置

将PaddleClas-2.4.0/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml配置文件中的内容改成以下内容

# global configs
Global:
  checkpoints: null
  pretrained_model: null
  output_dir: ./output/
  device: gpu
  save_interval: 5
  eval_during_train: True
  eval_interval: 5
  epochs: 300
  print_batch_step: 10
  use_visualdl: False
  # used for static mode and model export
  image_shape: [3, 224, 224]
  save_inference_dir: ./inference
  # training model under @to_static
  to_static: False
  use_multilabel: True

# model architecture
Arch:
  name: SwinTransformer_tiny_patch4_window7_224
  class_num: 33
  pretrained: True
 
# loss function config for traing/eval process
Loss:
  Train:
    - MultiLabelLoss:
        weight: 1.0
  Eval:
    - MultiLabelLoss:
        weight: 1.0


Optimizer:
  name: AdamW
  beta1: 0.9
  beta2: 0.999
  epsilon: 1e-8
  weight_decay: 0.05
  no_weight_decay_name: absolute_pos_embed relative_position_bias_table .bias norm 
  one_dim_param_no_weight_decay: True
  lr:
    # for 8 cards
    name: Cosine
    learning_rate: 1e-3
    eta_min: 2e-5
    warmup_epoch: 20
    warmup_start_lr: 2e-6


# data loader for train and eval
DataLoader:
  Train:
    dataset:
      name: MultiLabelDataset
      image_root: ./dataset/NUS-WIDE-SCENE/NUS-SCENE-dataset/images/
      cls_label_path: ./dataset/NUS-WIDE-SCENE/NUS-SCENE-dataset/multilabel_train_list.txt
      transform_ops:
        - DecodeImage:
            to_rgb: True
            channel_first: False
        - RandCropImage:
            size: 224
            interpolation: bicubic
            backend: pil
        - RandFlipImage:
            flip_code: 1
        - TimmAutoAugment:
            config_str: rand-m9-mstd0.5-inc1
            interpolation: bicubic
            img_size: 224
        - NormalizeImage:
            scale: 1.0/255.0
            mean: [0.485, 0.456, 0.406]
            std: [0.229, 0.224, 0.225]
            order: ''
        - RandomErasing:
            EPSILON: 0.25
            sl: 0.02
            sh: 1.0/3.0
            r1: 0.3
            attempt: 10
            use_log_aspect: True
            mode: pixel
      # batch_transform_ops:
      #   - OpSampler:
      #       MixupOperator:
      #         alpha: 0.8
      #         prob: 0.5
      #       CutmixOperator:
      #         alpha: 1.0
      #         prob: 0.5

    sampler:
      name: DistributedBatchSampler
      batch_size: 128
      drop_last: False
      shuffle: True
    loader:
      num_workers: 4
      use_shared_memory: True

  Eval:
    dataset: 
      name: MultiLabelDataset
      image_root: ./dataset/NUS-WIDE-SCENE/NUS-SCENE-dataset/images/
      cls_label_path: ./dataset/NUS-WIDE-SCENE/NUS-SCENE-dataset/multilabel_test_list.txt
      transform_ops:
        - DecodeImage:
            to_rgb: True
            channel_first: False
        - ResizeImage:
            resize_short: 256
            interpolation: bicubic
            backend: pil
        - CropImage:
            size: 224
        - NormalizeImage:
            scale: 1.0/255.0
            mean: [0.485, 0.456, 0.406]
            std: [0.229, 0.224, 0.225]
            order: ''
    sampler:
      name: DistributedBatchSampler
      batch_size: 128
      drop_last: False
      shuffle: False
    loader:
      num_workers: 4
      use_shared_memory: True

Infer:
  infer_imgs: ./deploy/images/0517_2715693311.jpg
  batch_size: 10
  transforms:
    - DecodeImage:
        to_rgb: True
        channel_first: False
    - ResizeImage:
        resize_short: 256
        interpolation: bicubic
        backend: pil
    - CropImage:
        size: 224
    - NormalizeImage:
        scale: 1.0/255.0
        mean: [0.485, 0.456, 0.406]
        std: [0.229, 0.224, 0.225]
        order: ''
    - ToCHWImage:
  PostProcess:
    name: MultiLabelTopk
    topk: 5
    class_id_map_file: None

Metric:
  Train:
    - HammingDistance:
    - AccuracyScore:
  Eval:
    - HammingDistance:
    - AccuracyScore:

!cp /home/aistudio/SwinTransformer_tiny_patch4_window7_224.yaml ~/PaddleClas-2.4.0/ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml

3.开始训练

%cd ~/PaddleClas-2.4.0/
!python3  tools/train.py \
        -c ./ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml

五、模型评估

!python3 tools/eval.py \
    -c ./ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml \
    -o Arch.pretrained="./output/SwinTransformer_tiny_patch4_window7_224/latest"

六、模型预测

通过预测，图像

!python3 tools/infer.py \
    -c ./ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml \
    -o Arch.pretrained="./output/SwinTransformer_tiny_patch4_window7_224/latest"

七、基于预测引擎预测

1.导出 inference model

!python3 tools/export_model.py \
    -c ./ppcls/configs/ImageNet/SwinTransformer/SwinTransformer_tiny_patch4_window7_224.yaml \
    -o Arch.pretrained="./output/SwinTransformer_tiny_patch4_window7_224/latest"

inference model 的路径默认在当前路径下 ./inference

%cd ~/PaddleClas-2.4.0
!ls  ./inference -l

2 基于预测引擎预测

首先进入 deploy 目录
通过预测引擎推理预测

将PaddleClas-2.4.0/deploy/configs/inference_cls_multilabel.yaml配置文件修改成如下内容

Global:
  infer_imgs: "./images/0517_2715693311.jpg"
  inference_model_dir: "../inference/"
  batch_size: 1
  use_gpu: True
  enable_mkldnn: False
  cpu_num_threads: 10
  enable_benchmark: True
  use_fp16: False
  ir_optim: True
  use_tensorrt: False
  gpu_mem: 8000
  enable_profile: False

PreProcess:
  transform_ops:
    - ResizeImage:
        resize_short: 256
    - CropImage:
        size: 224
    - NormalizeImage:
        scale: 0.00392157
        mean: [0.485, 0.456, 0.406]
        std: [0.229, 0.224, 0.225]
        order: ''
        channel_num: 3
    - ToCHWImage:
    
PostProcess:
  main_indicator: MultiLabelTopk
  MultiLabelTopk:
    topk: 5
    class_id_map_file: None
  SavePreLabel:
    save_dir: ./pre_label/

!cp /home/aistudio/inference_cls_multilabel.yaml ~/inference_cls_multilabel.yaml

%cd ~/PaddleClas-2.4.0/deploy
!python3 python/predict_cls.py \
     -c ./configs/inference_cls_multilabel.yaml

八、总结

多标签图像分类在日常生活中很常见，例如百度网盘AI大赛——版式分析场景比赛https://aistudio.baidu.com/aistudio/competition/detail/850/0/leaderboard等。

上述比赛中，可以利用更新的模型来训练多标签分类，将取得更好的结果。

此文章为搬运
原项目链接

百度飞桨AI Studio社区

学大模型，用大模型上飞桨星河社区！每天8点V100G算力免费领！免费领取ERNIE 4.0 100w Token >>>

更多推荐

Python调用文心一言的API

百度飞桨星河社区

文心一言API调用（python）

百度飞桨星河社区

[PaddleSpeech 原神] 音色克隆之胡桃

[PaddleSpeech] 音色克隆之原神角色，项目代码已完整放在aistudio,地址如下：快去体验吧，可以直接运行。

百度飞桨星河社区

所有评论(0)

查看更多评论

AI Studio

@m0_63642362

已为社区贡献1438条内容