init

ConvNeXt_v2-v2_ep90.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c824bc8322a1bc76cfd1e56c94fc76bf380c7d2890148b3a5d5e8a3d27d47d3
+size 449531965

app.py

@@ -0,0 +1,78 @@
+
+
+import os
+import json
+from mmpretrain import ImageClassificationInferencer
+import torch
+import gradio as gr
+
+config = 'convnext-v2-tiny_32xb32_in1k-384px.py'
+checkpoint = 'ConvNeXt_v2-v2_ep90.pth'
+
+
+inferencer = ImageClassificationInferencer(model=config, pretrained=checkpoint, device= "cuda" if torch.cuda.is_available() else "cpu")
+
+def single_image_classifier(image):
+    inf_result = inferencer(image)[0]
+    label = inf_result['pred_class']
+    score = inf_result['pred_score']
+    if label == "not_western":
+        another_label = "western"
+        another_score = 1 - score
+    else:
+        another_label = "not_western"
+        another_score = 1 - score
+    return {label : score,another_label:another_score}
+
+def batch_process(path,is_pred_score):
+    result={}
+    try:
+        for root, dirs, files in os.walk(path):
+            for file in files:
+                if file.lower().endswith(('.png', '.jpg','jpeg')):
+                    inf_result = inferencer(os.path.join(root, file))[0]
+                    print(result,os.path.join(root, file))
+                    if is_pred_score == True:
+                        result[os.path.join(root, file)]= [{'pred_class' : inf_result['pred_class']},{'pred_score' : inf_result['pred_score']}]
+                    else:
+                        result[os.path.join(root, file)]= [{'pred_class' : inf_result['pred_class']}]
+        with open(path+ "/" + "predict_result.json", "w") as file:
+            json.dump(result, file, ensure_ascii=False,indent=2)
+        return "sucess"
+    except:
+        return "failed"
+
+with gr.Blocks() as demo:
+    gr.Markdown("# Western anime images classification")
+    gr.Markdown("A classification using mmpretrain trained to classify western images based on ConvNeXtV2-tiny.Used for classifying anime images based on whether they are in the Western style.\n\n"
+                "The inference script: https://huggingface.co/TLME/western-classification"
+                )
+    with gr.Tab("Single image"):
+            input_img = gr.Image(source='upload')
+            
+            
+            output_label =gr.Label(label="Predict result")
+            examples_imgs = ["./testimg2/1.jpg","./testimg2/2.jpg","./testimg2/3.jpg","./testimg2/4.jpg","./testimg/1.jpg","./testimg/2.jpg","./testimg/3.jpg","./testimg/4.jpg"]
+            
+            button = gr.Button("Submit",variant="primary")
+            
+            button.click(single_image_classifier,inputs= input_img,outputs= output_label)
+            gr.Examples(examples= examples_imgs ,inputs = input_img , outputs= output_label,fn=single_image_classifier)
+            
+    with gr.Tab("Batch process"):
+        with gr.Row(label ='Input path'):
+            with gr.Column():
+                input_path = gr.Textbox(label="input your images folder")
+                is_pred_score = gr.Checkbox(value = True, label="Output pred_score")
+            
+        output_msg = gr.Textbox(label="Message")
+
+        buttom =gr.Button("Process",variant="primary")
+        buttom.click(batch_process, inputs= [input_path,is_pred_score] , outputs=output_msg)
+        
+if __name__ == "__main__":
+    demo.queue(concurrency_count=4)
+    demo.launch()
+    
+    
+

convnext-v2-tiny_32xb32_in1k-384px.py

@@ -0,0 +1,180 @@
+auto_scale_lr = dict(base_batch_size=96)
+custom_hooks = [
+    dict(momentum=0.0001, priority='ABOVE_NORMAL', type='EMAHook'),
+]
+data_preprocessor = dict(
+    mean=[
+        123.675,
+        116.28,
+        103.53,
+    ],
+    num_classes=2,
+    std=[
+        58.395,
+        57.12,
+        57.375,
+    ],
+    to_rgb=True)
+dataset_type = 'CustomDataset'
+default_hooks = dict(
+    checkpoint=dict(interval=2, type='CheckpointHook'),
+    logger=dict(interval=100, type='LoggerHook'),
+    param_scheduler=dict(type='ParamSchedulerHook'),
+    sampler_seed=dict(type='DistSamplerSeedHook'),
+    timer=dict(type='IterTimerHook'),
+    visualization=dict(
+        enable=True,
+        interval=1,
+        out_dir=None,
+        type='VisualizationHook',
+        wait_time=2))
+default_scope = 'mmpretrain'
+env_cfg = dict(
+    cudnn_benchmark=False,
+    dist_cfg=dict(backend='nccl'),
+    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0))
+launcher = 'none'
+load_from = './ConvNeXt_v2-v2_ep90.pth'
+log_level = 'INFO'
+model = dict(
+    backbone=dict(
+        arch='tiny',
+        drop_path_rate=0.5,
+        layer_scale_init_value=0.0,
+        type='ConvNeXt',
+        use_grn=True),
+    head=dict(
+        in_channels=768,
+        init_cfg=None,
+        loss=dict(label_smooth_val=0.2, type='LabelSmoothLoss'),
+        num_classes=2,
+        type='LinearClsHead'),
+    init_cfg=dict(
+        bias=0.0, layer=[
+            'Conv2d',
+            'Linear',
+        ], std=0.02, type='TruncNormal'),
+    train_cfg=dict(augments=[
+        dict(alpha=0.8, type='Mixup'),
+        dict(alpha=1.0, type='CutMix'),
+    ]),
+    type='ImageClassifier')
+optim_wrapper = dict(
+    accumulative_counts=3,
+    clip_grad=None,
+    loss_scale='dynamic',
+    optimizer=dict(
+        betas=(
+            0.9,
+            0.999,
+        ),
+        eps=1e-08,
+        lr=0.00032,
+        type='AdamW',
+        weight_decay=0.05),
+    paramwise_cfg=dict(
+        bias_decay_mult=0.0,
+        custom_keys=dict({
+            '.absolute_pos_embed': dict(decay_mult=0.0),
+            '.relative_position_bias_table': dict(decay_mult=0.0)
+        }),
+        flat_decay_mult=0.0,
+        norm_decay_mult=0.0),
+    type='AmpOptimWrapper')
+param_scheduler = [
+    dict(
+        by_epoch=True,
+        convert_to_iter_based=True,
+        end=2,
+        start_factor=0.001,
+        type='LinearLR'),
+    dict(begin=2, by_epoch=True, eta_min=8e-05, type='CosineAnnealingLR'),
+]
+randomness = dict(deterministic=False, seed=None)
+resume = False
+test_cfg = dict()
+test_dataloader = dict(
+    batch_size=16,
+    collate_fn=dict(type='default_collate'),
+    dataset=dict(
+        data_root='./testimgs',
+        pipeline=[
+            dict(type='LoadImageFromFile'),
+            dict(
+                backend='pillow',
+                interpolation='bicubic',
+                scale=384,
+                type='Resize'),
+            dict(type='PackInputs'),
+        ],
+        type='CustomDataset'),
+    num_workers=5,
+    persistent_workers=True,
+    pin_memory=True,
+    sampler=dict(shuffle=False, type='DefaultSampler'))
+test_evaluator = dict(topk=(1, ), type='Accuracy')
+test_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(backend='pillow', interpolation='bicubic', scale=384, type='Resize'),
+    dict(type='PackInputs'),
+]
+train_cfg = dict(by_epoch=True, max_epochs=120, val_interval=1)
+train_dataloader = dict(
+    batch_size=32,
+    collate_fn=dict(type='default_collate'),
+    dataset=dict(
+        data_root='./procset',
+        pipeline=[
+            dict(type='LoadImageFromFile'),
+            dict(
+                backend='pillow',
+                interpolation='bicubic',
+                scale=384,
+                type='RandomResizedCrop'),
+            dict(direction='horizontal', prob=0.5, type='RandomFlip'),
+            dict(type='PackInputs'),
+        ],
+        type='CustomDataset'),
+    num_workers=5,
+    persistent_workers=True,
+    pin_memory=True,
+    sampler=dict(shuffle=True, type='DefaultSampler'))
+train_pipeline = [
+    dict(type='LoadImageFromFile'),
+    dict(
+        backend='pillow',
+        interpolation='bicubic',
+        scale=384,
+        type='RandomResizedCrop'),
+    dict(direction='horizontal', prob=0.5, type='RandomFlip'),
+    dict(type='PackInputs'),
+]
+val_cfg = dict()
+val_dataloader = dict(
+    batch_size=16,
+    collate_fn=dict(type='default_collate'),
+    dataset=dict(
+        data_root='./valset',
+        pipeline=[
+            dict(type='LoadImageFromFile'),
+            dict(
+                backend='pillow',
+                interpolation='bicubic',
+                scale=384,
+                type='Resize'),
+            dict(type='PackInputs'),
+        ],
+        type='CustomDataset'),
+    num_workers=5,
+    persistent_workers=True,
+    pin_memory=True,
+    sampler=dict(shuffle=False, type='DefaultSampler'))
+val_evaluator = dict(topk=(1, ), type='Accuracy')
+vis_backends = [
+    dict(type='LocalVisBackend'),
+]
+visualizer = dict(
+    type='UniversalVisualizer', vis_backends=[
+        dict(type='LocalVisBackend'),
+    ])
+work_dir = './work_dirs\\convnext-v2-tiny_32xb32_in1k-384px'

infer.py

@@ -0,0 +1,27 @@
+
+
+import os
+import json
+from mmpretrain import ImageClassificationInferencer
+
+
+path = './testimg/'
+config = 'convnext-v2-tiny_32xb32_in1k-384px.py'
+checkpoint = 'ConvNeXt_v2-v2_ep90.pth'
+
+
+inferencer = ImageClassificationInferencer(model=config, pretrained=checkpoint, device='cuda')
+
+result={}
+
+for root, dirs, files in os.walk(path):
+    for file in files:
+        if file.lower().endswith(('.png', '.jpg','jpeg')):
+            # print(os.path.join(root, file))
+            inf_result = inferencer(os.path.join(root, file))[0]
+            # print(result['pred_class'])
+            print(result,os.path.join(root, file))
+            result[os.path.join(root, file)]= [{'pred_class' : inf_result['pred_class']},{'pred_score' : inf_result['pred_score']}]
+
+with open(path + "predict_result.json", "w") as file:
+    json.dump(result, file, ensure_ascii=False,indent=2)

requirements.txt

@@ -0,0 +1,5 @@
+openmim
+mmcv>=2.0.0,<2.1.0
+mmengine>=0.8.3,<1.0.0
+mmpretrain
+torch

testimg2/predict_result.json

@@ -0,0 +1,22 @@
+{
+  "F:\\aethe_checker\\western-classification\\testimg2\\1.jpg": [
+    {
+      "pred_class": "western"
+    }
+  ],
+  "F:\\aethe_checker\\western-classification\\testimg2\\2.jpg": [
+    {
+      "pred_class": "western"
+    }
+  ],
+  "F:\\aethe_checker\\western-classification\\testimg2\\3.jpg": [
+    {
+      "pred_class": "western"
+    }
+  ],
+  "F:\\aethe_checker\\western-classification\\testimg2\\4.jpg": [
+    {
+      "pred_class": "western"
+    }
+  ]
+}