add app

.gitignore

@@ -0,0 +1 @@
+__pycache__

app.py

@@ -0,0 +1,85 @@
+import gradio as gr
+from ultralytics import YOLO
+
+import os
+from pathlib import Path
+
+model_paths = {
+    'YOLOv8n': '/mnt/d/Users/E1005164/articles/deadwood-yolov8/codes/runs_pretrained/yolov8n_Adam',
+    'YOLOv8s': '/mnt/d/Users/E1005164/articles/deadwood-yolov8/codes/runs_pretrained/yolov8s_Adam',
+    'YOLOv8m': '/mnt/d/Users/E1005164/articles/deadwood-yolov8/codes/runs_pretrained/yolov8m_Adam',
+    'YOLOv8l': '/mnt/d/Users/E1005164/articles/deadwood-yolov8/codes/runs_pretrained/yolov8l_Adam',
+    'YOLOv8x': '/mnt/d/Users/E1005164/articles/deadwood-yolov8/codes/runs_pretrained/yolov8x_Adam',
+}
+
+def run_models(
+        im:gr.Image=None,
+        model_type:gr.Dropdown='YOLOv8n',
+        conf_thr:gr.Slider=0.25
+):
+    
+    hp_model = f'{model_paths[model_type]}_hp/weights/best.pt'
+    model = YOLO(hp_model)
+    model.to(device='cpu')
+    hp_result = model(im[:,:,::-1], conf=conf_thr)
+    hp_im = hp_result[0].plot()
+
+    spk_model = f'{model_paths[model_type]}_spk/weights/best.pt'
+    model = YOLO(spk_model)
+    model.to(device='cpu')
+
+    spk_result = model(im[:,:,::-1], conf=conf_thr)
+    spk_im = spk_result[0].plot()
+
+    both_model = f'{model_paths[model_type]}_both/weights/best.pt'
+    model = YOLO(both_model)
+    model.to(device='cpu')
+
+    both_result = model(im[:,:,::-1], conf=conf_thr)
+    both_im = both_result[0].plot()
+    return [
+        (hp_im[:,:,::-1], 'HP'),
+        (spk_im[:,:,::-1], 'SPK'),
+        (both_im[:,:,::-1], 'HP+SPK')
+    ]
+
+ex_dir = Path('examples')
+
+loc = gr.Textbox(label='Location')
+
+desc_str = """
+Demo application for YOLOv8 models for deadwood segmentation from RGB UAV imagery. Results are shown on three different models: HP is trained only with data from Hiidenportti, 
+SPK only with data from Sudenpesänkangas and HP+SPK is trained with both sites.
+"""
+
+with gr.Blocks() as demo:
+    with gr.Row():
+        gr.Markdown(desc_str)
+    with gr.Row():
+        with gr.Column(2):
+            inp = gr.Image(label='Input image', sources='upload')
+        with gr.Column(1):
+            ex_list = [[ex_dir/i, i.split('_')[0]] for i in os.listdir(ex_dir)]
+            ex = gr.Examples(ex_list, inputs=[inp, loc], label=None,
+                             cache_examples=False, examples_per_page=5)
+        with gr.Column(1):
+            loc.render()
+            model = gr.Dropdown([
+                        'YOLOv8n', 
+                        'YOLOv8s', 
+                        'YOLOv8m', 
+                        'YOLOv8l', 
+                        'YOLOv8x'
+                        ],
+                        value='YOLOv8n', label='Model')
+            conf = gr.Slider(minimum=0.0, maximum=1.0, value=0.25, step=0.05, label='Confidence Threshold')
+            btn = gr.Button()
+    with gr.Row():
+        with gr.Column():
+            gallery = gr.Gallery(
+                label='Predictions', show_label=True, elem_id='gallery',
+                columns=[3], rows=[1], object_fit='contain', interactive=False
+            ) 
+    btn.click(run_models, [inp, model, conf], gallery)
+
+if __name__ == '__main__': demo.launch(share=False)