cella110n commited on
Commit
e6e216d
·
verified ·
1 Parent(s): 436ac8e

Upload 2 files

Browse files
Files changed (2) hide show
  1. README.md +28 -28
  2. app.py +7 -7
README.md CHANGED
@@ -1,29 +1,29 @@
1
- ---
2
- title: CL EVA02 LoRA ONNX Tagger
3
- emoji: 🖼️
4
- colorFrom: blue
5
- colorTo: green
6
- sdk: gradio
7
- sdk_version: 4.43.0 # requirements.txt と合わせるか確認
8
- app_file: app.py
9
- # license: apache-2.0 # または適切なライセンス
10
- # Pinned Hardware: T4 small (GPU) or CPU upgrade (CPU)
11
- # pinned: false # 必要に応じてTrueに
12
- # hardware: cpu-upgrade # or cuda-t4-small
13
- # hf_token: YOUR_HF_TOKEN # Use secrets instead!
14
- ---
15
-
16
- # CL EVA02 LoRA ONNX Tagger
17
-
18
- This Space demonstrates image tagging using a fine-tuned WD EVA02 model (converted to ONNX format).
19
-
20
- **How to Use:**
21
- 1. Upload an image using the upload button.
22
- 2. Alternatively, paste an image URL into the browser (experimental paste handling).
23
- 3. Adjust the tag thresholds if needed.
24
- 4. Choose the output mode (Tags only or include visualization).
25
- 5. Click the "Predict" button.
26
-
27
- **Note:**
28
- - This Space uses a model from a **private** repository (`celstk/wd-eva02-lora-onnx`). You might need to duplicate this space and add your Hugging Face token (`HF_TOKEN`) to the Space secrets to allow downloading the model files.
29
  - Image pasting behavior might vary across browsers.
 
1
+ ---
2
+ title: CL EVA02 LoRA ONNX Tagger
3
+ emoji: 🖼️
4
+ colorFrom: blue
5
+ colorTo: green
6
+ sdk: gradio
7
+ sdk_version: 4.43.0 # requirements.txt と合わせるか確認
8
+ app_file: app.py
9
+ # license: apache-2.0 # または適切なライセンス
10
+ # Pinned Hardware: T4 small (GPU) or CPU upgrade (CPU)
11
+ # pinned: false # 必要に応じてTrueに
12
+ # hardware: cpu-upgrade # or cuda-t4-small
13
+ # hf_token: YOUR_HF_TOKEN # Use secrets instead!
14
+ ---
15
+
16
+ # WD EVA02 LoRA ONNX Tagger
17
+
18
+ This Space demonstrates image tagging using a fine-tuned WD EVA02 model (converted to ONNX format).
19
+
20
+ **How to Use:**
21
+ 1. Upload an image using the upload button.
22
+ 2. Alternatively, paste an image URL into the browser (experimental paste handling).
23
+ 3. Adjust the tag thresholds if needed.
24
+ 4. Choose the output mode (Tags only or include visualization).
25
+ 5. Click the "Predict" button.
26
+
27
+ **Note:**
28
+ - This Space uses a model from a **private** repository (`celstk/wd-eva02-lora-onnx`). You might need to duplicate this space and add your Hugging Face token (`HF_TOKEN`) to the Space secrets to allow downloading the model files.
29
  - Image pasting behavior might vary across browsers.
app.py CHANGED
@@ -199,14 +199,14 @@ def visualize_predictions(image: Image.Image, predictions: Dict, threshold: floa
199
 
200
  # --- Plotting Setup ---
201
  plt.rcParams['font.family'] = 'DejaVu Sans' # Ensure font compatibility
202
- fig = plt.figure(figsize=(20, 12), dpi=100)
203
  gs = fig.add_gridspec(1, 2, width_ratios=[1.2, 1])
204
 
205
  # Left side: Image
206
- ax_img = fig.add_subplot(gs[0, 0])
207
- ax_img.imshow(image)
208
- ax_img.set_title("Original Image")
209
- ax_img.axis('off')
210
 
211
  # Right side: Tags
212
  ax_tags = fig.add_subplot(gs[0, 1])
@@ -288,7 +288,7 @@ def visualize_predictions(image: Image.Image, predictions: Dict, threshold: floa
288
  REPO_ID = "cella110n/cl_tagger"
289
  # Use the specified ONNX model filename
290
  ONNX_FILENAME = "cl_eva02_tagger_v1_250426/model.onnx"
291
- # Keep the previously used tag mapping filename
292
  TAG_MAPPING_FILENAME = "cl_eva02_tagger_v1_250426/tag_mapping.json"
293
  CACHE_DIR = "./model_cache"
294
 
@@ -467,7 +467,7 @@ footer { display: none !important; }
467
 
468
  with gr.Blocks(css=css) as demo:
469
  gr.Markdown("# CL EVA02 ONNX Tagger")
470
- gr.Markdown("Upload an image or paste an image URL to predict tags using the WD EVA02 Tagger model (ONNX), fine-tuned from [SmilingWolf/wd-eva02-large-tagger-v3](https://huggingface.co/SmilingWolf/wd-eva02-large-tagger-v3).")
471
  with gr.Row():
472
  with gr.Column(scale=1):
473
  image_input = gr.Image(type="pil", label="Input Image", elem_id="input-image")
 
199
 
200
  # --- Plotting Setup ---
201
  plt.rcParams['font.family'] = 'DejaVu Sans' # Ensure font compatibility
202
+ fig = plt.figure(figsize=(12, 20), dpi=100)
203
  gs = fig.add_gridspec(1, 2, width_ratios=[1.2, 1])
204
 
205
  # Left side: Image
206
+ # ax_img = fig.add_subplot(gs[0, 0])
207
+ # ax_img.imshow(image)
208
+ # ax_img.set_title("Original Image")
209
+ # ax_img.axis('off')
210
 
211
  # Right side: Tags
212
  ax_tags = fig.add_subplot(gs[0, 1])
 
288
  REPO_ID = "cella110n/cl_tagger"
289
  # Use the specified ONNX model filename
290
  ONNX_FILENAME = "cl_eva02_tagger_v1_250426/model.onnx"
291
+ # Correct the tag mapping path to match the ONNX model's directory
292
  TAG_MAPPING_FILENAME = "cl_eva02_tagger_v1_250426/tag_mapping.json"
293
  CACHE_DIR = "./model_cache"
294
 
 
467
 
468
  with gr.Blocks(css=css) as demo:
469
  gr.Markdown("# CL EVA02 ONNX Tagger")
470
+ gr.Markdown("Upload an image or paste an image URL to predict tags using the CL EVA02 Tagger model (ONNX), fine-tuned from [SmilingWolf/wd-eva02-large-tagger-v3](https://huggingface.co/SmilingWolf/wd-eva02-large-tagger-v3).")
471
  with gr.Row():
472
  with gr.Column(scale=1):
473
  image_input = gr.Image(type="pil", label="Input Image", elem_id="input-image")