First commit trying to make this run

.gitignore

@@ -0,0 +1 @@
+venv

app.py

@@ -0,0 +1,200 @@
+import gradio as gr
+import spaces
+from huggingface_hub import hf_hub_download
+import os
+
+# make sure you have the following dependencies
+import torch
+import numpy as np
+from models.common import DetectMultiBackend
+from utils.general import non_max_suppression, scale_boxes
+from utils.torch_utils import select_device, smart_inference_mode
+from utils.augmentations import letterbox
+import PIL.Image
+
+#@smart_inference_mode()
+@spaces.GPU
+def yolov9_inference(img_path, model_id='YOLOv9-S_X_LOCO-converted.pt', img_size=640, conf_thres=0.1, iou_thres=0.4):
+    """
+    Load a YOLOv9 model, configure it, perform inference on an image, and optionally adjust 
+    the input size and apply test time augmentation.
+    
+    :param model_path: Path to the YOLOv9 model file.
+    :param conf_threshold: Confidence threshold for NMS.
+    :param iou_threshold: IoU threshold for NMS.
+    :param img_path: Path to the image file.
+    :param size: Optional, input size for inference.
+    :return: A tuple containing the detections (boxes, scores, categories) and the results object for further actions like displaying.
+    """
+    
+    # Load the model
+    model_path = download_models(model_id)
+
+    # Initialize
+    device = select_device('0')
+    model = DetectMultiBackend(model_path, device="0", fp16=False, data='data/coco.yaml')
+    stride, names, pt = model.stride, model.names, model.pt
+
+    # Load image
+    img = np.array(PIL.Image.open(img_path))
+    img = letterbox(img0, img_size, stride=stride, auto=True)[0]
+    img = img[:, :, ::-1].transpose(2, 0, 1)
+    img = np.ascontiguousarray(img)
+    img = torch.from_numpy(img).to(device).float()
+    img /= 255.0
+    if img.ndimension() == 3:
+        img = img.unsqueeze(0)
+
+    # Inference
+    results = model(img, augment=False, visualize=False)
+
+    # Apply NMS
+    results = non_max_suppression(results[0][0], conf_thres, iou_thres, classes=None, max_det=1000)
+
+    output = results.render()
+
+    return output[0]
+
+
+
+def download_models(model_id):
+    hf_hub_download("KoniHD/LOCO-Detection", filename=f"{model_id}", local_dir=f"./",
+                    token=os.getenv("HF_TOKEN"))
+    return f"./{model_id}"
+
+# @spaces.GPU
+# def yolov9_inference(img_path, model_id, image_size, conf_threshold, iou_threshold):
+#     """
+#     Load a YOLOv9 model, configure it, perform inference on an image, and optionally adjust 
+#     the input size and apply test time augmentation.
+    
+#     :param model_path: Path to the YOLOv9 model file.
+#     :param conf_threshold: Confidence threshold for NMS.
+#     :param iou_threshold: IoU threshold for NMS.
+#     :param img_path: Path to the image file.
+#     :param size: Optional, input size for inference.
+#     :return: A tuple containing the detections (boxes, scores, categories) and the results object for further actions like displaying.
+#     """
+#     # Import YOLOv9
+#     import yolov9
+    
+#     # Load the model
+#     model_path = download_models(model_id)
+#     model = yolov9.load(model_path, device="cuda:0")
+    
+#     # Set model parameters
+#     model.conf = conf_threshold
+#     model.iou = iou_threshold
+    
+#     # Perform inference
+#     results = model(img_path, size=image_size)
+
+#     # Optionally, show detection bounding boxes on image
+#     output = results.render()
+    
+#     return output[0]
+
+
+def app():
+    with gr.Blocks():
+        with gr.Row():
+            with gr.Column():
+                img_path = gr.Image(type="filepath", label="Image")
+                model_path = gr.Dropdown(
+                    label="Model",
+                    choices=[
+                        "YOLOv9-S_X_LOCO-converted.pt",
+                        "YOLOv9_S_X_LOCO.pt",
+                        "YOLOv9-E_X_LOCO-converted.pt",
+                        "YOLOv9_E_X_LOCO.pt",
+                    ],
+                    value="YOLOv9-S_X_LOCO-converted.pt",
+                )
+                image_size = gr.Slider(
+                    label="Image Size",
+                    minimum=320,
+                    maximum=1280,
+                    step=32,
+                    value=640,
+                )
+                conf_threshold = gr.Slider(
+                    label="Confidence Threshold",
+                    minimum=0.1,
+                    maximum=1.0,
+                    step=0.1,
+                    value=0.4,
+                )
+                iou_threshold = gr.Slider(
+                    label="IoU Threshold",
+                    minimum=0.1,
+                    maximum=1.0,
+                    step=0.1,
+                    value=0.5,
+                )
+                yolov9_infer = gr.Button(value="Inference")
+
+            with gr.Column():
+                output_numpy = gr.Image(type="numpy",label="Output")
+
+        yolov9_infer.click(
+            fn=yolov9_inference,
+            inputs=[
+                img_path,
+                model_path,
+                image_size,
+                conf_threshold,
+                iou_threshold,
+            ],
+            outputs=[output_numpy],
+        )
+        
+        gr.Examples(
+            examples=[
+                [
+                    "data/zidane.jpg",
+                    "YOLOv9-S_X_LOCO-converted.pt",
+                    640,
+                    0.4,
+                    0.5,
+                ],
+                [
+                    "data/huggingface.jpg",
+                    "YOLOv9-E_X_LOCO-converted.pt",
+                    640,
+                    0.4,
+                    0.5,
+                ],
+            ],
+            fn=yolov9_inference,
+            inputs=[
+                img_path,
+                model_path,
+                image_size,
+                conf_threshold,
+                iou_threshold,
+            ],
+            outputs=[output_numpy],
+            cache_examples=True,
+        )
+
+
+gradio_app = gr.Blocks()
+with gradio_app:
+    gr.HTML(
+        """
+    <h1 style='text-align: center'>
+    YOLOv9: Learning What You Want to Learn Using Programmable Gradient Information
+    </h1>
+    """)
+    gr.HTML(
+        """
+        <h3 style='text-align: center'>
+        Follow me for more!
+        <a href='https://twitter.com/konihd_7' target='_blank'>Twitter</a> | <a href='https://github.com/KoniHD' target='_blank'>Github</a> | <a href='https://www.linkedin.com/in/konstantin-zeck/' target='_blank'>Linkedin</a>  | <a href='https://www.huggingface.co/KoniHD/' target='_blank'>HuggingFace</a>
+        </h3>
+        """)
+    with gr.Row():
+        with gr.Column():
+            app()
+
+gradio_app.launch(debug=True)

requirements.txt

@@ -1,8 +1,6 @@
 # requirements
 # Usage: pip install -r requirements.txt
 
-git+https://github.com/WongKinYiu/yolov9.git@main
-
 # Base ------------------------------------------------------------------------
 gitpython
 ipython
@@ -42,6 +40,8 @@ seaborn>=0.11.0
 
 # Deploy ----------------------------------------------------------------------
 # tritonclient[all]~=2.24.0
+git+https://github.com/KoniHD/yolov9.git@main#egg=yolov9
+huggingface_hub
 
 # Extras ----------------------------------------------------------------------
 # mss