Upload 2 files

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+raccoon-101.jpg filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,117 @@
+from huggingface_hub import from_pretrained_fastai
+import gradio as gr
+from fastai.vision.all import *
+from PIL import Image, ImageDraw, ImageFont
+import numpy as np
+
+repo_id = "magomerob/yolo_finetuned_raccoons"
+learner = from_pretrained_fastai(repo_id)
+labels = learner.dls.vocab
+
+
+def _to_pil(x):
+    if isinstance(x, Image.Image):
+        return x.convert("RGB")
+    if isinstance(x, np.ndarray):
+        return Image.fromarray(x).convert("RGB")
+    return PILImage.create(x).convert("RGB")
+
+
+def _get_detections(pred, labels_vocab):
+    """
+    Returns: boxes_xyxy (list of [x1,y1,x2,y2]), names (list[str]), scores (list[float] or None)
+
+    Supports a few common formats:
+      - dict with keys like boxes/labels/scores
+      - tuple/list like (boxes, labels, scores)
+      - fastai-like (pred, pred_idx, probs) where pred holds detections
+    """
+    # fastai often returns (pred, pred_idx, probs)
+    if isinstance(pred, (tuple, list)) and len(pred) == 3:
+        pred = pred[0]
+
+    # dict-like output
+    if isinstance(pred, dict):
+        boxes = pred.get("boxes") or pred.get("bboxes") or pred.get("bbox")
+        lab = pred.get("labels") or pred.get("classes") or pred.get("label_ids")
+        scores = pred.get("scores") or pred.get("confs") or pred.get("confidences")
+        if boxes is None or lab is None:
+            raise ValueError(f"Unsupported dict output keys: {list(pred.keys())}")
+        boxes = np.asarray(boxes).tolist()
+        lab = np.asarray(lab).tolist()
+        scores = None if scores is None else np.asarray(scores).tolist()
+
+    # tuple/list like (boxes, labels, scores?)
+    elif isinstance(pred, (tuple, list)) and len(pred) >= 2:
+        boxes = np.asarray(pred[0]).tolist()
+        lab = np.asarray(pred[1]).tolist()
+        scores = None
+        if len(pred) >= 3 and pred[2] is not None:
+            try:
+                scores = np.asarray(pred[2]).tolist()
+            except Exception:
+                scores = None
+    else:
+        raise ValueError(f"Unsupported prediction type: {type(pred)}")
+
+    names = []
+    for x in lab:
+        try:
+            xi = int(x)
+            names.append(str(labels_vocab[xi]) if xi < len(labels_vocab) else str(xi))
+        except Exception:
+            names.append(str(x))
+
+    return boxes, names, scores
+
+
+def show_preds(input_image, display_label=True, display_bbox=True, detection_threshold=0.5):
+    if detection_threshold == 0:
+        detection_threshold = 0.5
+
+    img = _to_pil(input_image)
+    pred = learner.predict(img)
+
+    boxes, names, scores = _get_detections(pred, labels)
+
+    draw = ImageDraw.Draw(img)
+    try:
+        font = ImageFont.truetype("DejaVuSans.ttf", 16)
+    except Exception:
+        font = ImageFont.load_default()
+
+    for i, box in enumerate(boxes):
+        score = None if scores is None or i >= len(scores) else float(scores[i])
+        if score is not None and score < detection_threshold:
+            continue
+
+        x1, y1, x2, y2 = [int(round(v)) for v in box]
+
+        if display_bbox:
+            draw.rectangle([x1, y1, x2, y2], width=3)
+
+        if display_label:
+            label = names[i]
+            text = label + (f" {score:.2f}" if score is not None else "")
+            # small filled background for readability
+            tw, th = draw.textbbox((0, 0), text, font=font)[2:]
+            pad = 3
+            draw.rectangle([x1, max(0, y1 - th - 2 * pad), x1 + tw + 2 * pad, y1], fill="black")
+            draw.text((x1 + pad, max(0, y1 - th - pad)), text, font=font, fill="white")
+
+    return img
+
+
+display_chkbox_label = gr.Checkbox(label="Label", value=True)
+display_chkbox_box = gr.Checkbox(label="Box", value=True)
+detection_threshold_slider = gr.Slider(minimum=0, maximum=1, step=0.05, value=0.5, label="Detection Threshold")
+
+demo = gr.Interface(
+    fn=show_preds,
+    inputs=[gr.Image(type="numpy"), display_chkbox_label, display_chkbox_box, detection_threshold_slider],
+    outputs=gr.Image(type="pil"),
+    examples=[["raccoon-101.jpg", True, True, 0.5]],
+)
+
+if __name__ == "__main__":
+    demo.launch()