Upload 6 files

app.py

@@ -238,11 +238,11 @@ def draw_annotations(image_id):
 
 
 def add_annotation(image_id, cls, x, y, w, h):
-    if not image_id:return annotation_canvas_html(image_id),"Select an image first.",[]
-    if not cls:return annotation_canvas_html(image_id),"Select a class first.",[]
+    if not image_id:return annotation_preview_with_boxes(image_id),"Select an image first.",[]
+    if not cls:return annotation_preview_with_boxes(image_id),"Select a class first.",[]
     try:x,y,w,h=map(float,[x,y,w,h])
-    except:return annotation_canvas_html(image_id),"Enter box coordinates first.",[]
-    if w<=0 or h<=0:return annotation_canvas_html(image_id),"Box must have a width and height.",[]
+    except:return annotation_preview_with_boxes(image_id),"Enter box coordinates first.",[]
+    if w<=0 or h<=0:return annotation_preview_with_boxes(image_id),"Box must have a width and height.",[]
     data=load_dataset()
     item=next((z for z in data["images"] if z["id"]==image_id),None)
     if not item:return None,"Image not found.",[]
@@ -250,41 +250,72 @@ def add_annotation(image_id, cls, x, y, w, h):
     w=min(w,item["width"]-x); h=min(h,item["height"]-y)
     item.setdefault("annotations",[]).append({"class":cls,"box":[x,y,w,h]})
     save_dataset(data)
-    return annotation_canvas_html(image_id),f"Saved {cls}: [{x:.0f}, {y:.0f}, {w:.0f}, {h:.0f}]",item["annotations"]
+    return annotation_preview_with_boxes(image_id),f"Saved {cls}: [{x:.0f}, {y:.0f}, {w:.0f}, {h:.0f}]",item["annotations"]
 
 
 def remove_annotation(image_id,index):
-    if not image_id:return annotation_canvas_html(image_id),"Select an image first.",[]
+    if not image_id:return annotation_preview_with_boxes(image_id),"Select an image first.",[]
     data=load_dataset()
     item=next((z for z in data["images"] if z["id"]==image_id),None)
     if not item:return None,"Image not found.",[]
     try:idx=int(index)-1
-    except:return annotation_canvas_html(image_id),"Enter an annotation number.",item.get("annotations",[])
+    except:return annotation_preview_with_boxes(image_id),"Enter an annotation number.",item.get("annotations",[])
     anns=item.get("annotations",[])
-    if idx<0 or idx>=len(anns):return annotation_canvas_html(image_id),"Annotation number not found.",anns
+    if idx<0 or idx>=len(anns):return annotation_preview_with_boxes(image_id),"Annotation number not found.",anns
     deleted=anns.pop(idx);save_dataset(data)
-    return annotation_canvas_html(image_id),f"Deleted annotation {index}: {deleted['class']}",anns
+    return annotation_preview_with_boxes(image_id),f"Deleted annotation {index}: {deleted['class']}",anns
 
 
 def clear_annotations(image_id):
-    if not image_id:return annotation_canvas_html(image_id),"Select an image first.",[]
+    if not image_id:return annotation_preview_with_boxes(image_id),"Select an image first.",[]
     data=load_dataset()
     item=next((z for z in data["images"] if z["id"]==image_id),None)
     if not item:return None,"Image not found.",[]
     item["annotations"]=[];save_dataset(data)
-    return annotation_canvas_html(image_id),"Annotations cleared.",[]
+    return annotation_preview_with_boxes(image_id),"Annotations cleared.",[]
 
 
 def save_classes(text):
     classes = [x.strip() for x in (text or "").splitlines() if x.strip()]
     if not classes:
-        return "At least one class is required.", gr.update(choices=[]), dataset_status()
+        return "At least one class is required.", gr.update(choices=read_classes()), dataset_status()
     if len(set(classes)) != len(classes):
-        return "Classes must be unique.", gr.update(choices=CLASSES), dataset_status()
+        return "Classes must be unique.", gr.update(choices=read_classes()), dataset_status()
     CLASSES_FILE.write_text("\n".join(classes) + "\n", encoding="utf-8")
     data = load_dataset()
     save_dataset(data)
-    return f"Saved {len(classes)} classes.", gr.update(choices=CLASSES, value=classes[0]), dataset_status()
+    return f"Saved {len(classes)} classes.", gr.update(choices=classes, value=classes[0]), dataset_status()
+
+
+def handle_annotation_click(image_id, cls, click_state, evt: gr.SelectData):
+    """Use two clicks on the real Gradio image to define a box.
+    First click = top-left corner, second click = opposite corner.
+    This avoids the unreliable HTML canvas/script path and works in Gradio itself.
+    """
+    if not image_id:
+        return 0, 0, 0, 0, [], "Select an image first."
+    if not cls:
+        return 0, 0, 0, 0, [], "Select a class first."
+    data = load_dataset()
+    item = next((x for x in data["images"] if x["id"] == image_id), None)
+    if not item:
+        return 0, 0, 0, 0, [], "Image not found."
+    try:
+        point = evt.index
+        px, py = float(point[0]), float(point[1])
+    except Exception:
+        return 0, 0, 0, 0, click_state or [], "Could not read the image click position."
+    px = max(0, min(px, item["width"] - 1))
+    py = max(0, min(py, item["height"] - 1))
+    state = list(click_state or [])
+    if not state:
+        return round(px), round(py), 0, 0, [px, py], f"First corner: ({px:.0f}, {py:.0f}). Now click the opposite corner."
+    x0, y0 = state[:2]
+    x = min(x0, px); y = min(y0, py)
+    w = abs(px - x0); h = abs(py - y0)
+    if w < 2 or h < 2:
+        return round(x), round(y), 0, 0, [], "Box is too small. Click the first corner again."
+    return round(x), round(y), round(w), round(h), [], f"Box ready: [{x:.0f}, {y:.0f}, {w:.0f}, {h:.0f}] for {cls}. Click Save Box."
 
 
 def build_coco():
@@ -435,6 +466,7 @@ CSS = """
 .gradio-container { max-width: 1250px !important; }
 h1 { margin-bottom: 0.2rem !important; }
 .anno-wrap{width:100%}.anno-toolbar{display:flex;gap:14px;flex-wrap:wrap;padding:10px 12px;margin-bottom:8px;border-radius:10px;background:#20242a}.anno-toolbar span{opacity:.85}.anno-canvas-wrap{width:100%;overflow:auto;border:1px solid #555;border-radius:10px;background:#111;padding:8px}.anno-canvas-wrap canvas{display:block;max-width:none;cursor:crosshair;touch-action:none;margin:auto}.anno-help{padding:8px 2px;opacity:.75}.anno-empty{padding:50px;text-align:center;border:1px dashed #777;border-radius:10px}
+#annotation-image img { max-height: 650px !important; object-fit: contain !important; }
 .status { padding: 10px 14px; border-radius: 10px; }
 """
 
@@ -459,29 +491,35 @@ with gr.Blocks(title="Ice Cream Dataset + Counter") as demo:
         # Class dropdown is updated after the Annotate tab creates it.
 
     with gr.Tab("2 · Annotate"):
-        gr.Markdown("### See the real training image and annotate it")
-        gr.Markdown("The actual uploaded photo is displayed below. Select a class and enter the box coordinates in the original image pixels, then click **Save Box**. Saved boxes appear in red.")
+        gr.Markdown("### Annotate training images")
+        gr.Markdown("Select an image above, choose a class, then **click the first corner and click the opposite corner** of each object. The real uploaded image is shown below. Click **Save Box** after each box.")
         with gr.Row():
             with gr.Column(scale=3):
-                annotation_canvas = gr.HTML(value='<div class="anno-empty">Select an image.</div>', label="Training image")
-                editor_info = gr.Markdown()
+                annotation_image = gr.Image(value=None, type="pil", interactive=False, label="Training image", height=650, elem_id="annotation-image")
+                editor_info = gr.Markdown("Select an image from the Dataset tab.")
             with gr.Column(scale=1):
-                ann_class = gr.Dropdown(choices=read_classes(), value=(read_classes()[0] if read_classes() else None), label="Class")
-                x = gr.Number(label="X (left)", value=0, precision=0, elem_id="anno-x")
-                y = gr.Number(label="Y (top)", value=0, precision=0, elem_id="anno-y")
-                w = gr.Number(label="Width", value=0, precision=0, elem_id="anno-w")
-                h = gr.Number(label="Height", value=0, precision=0, elem_id="anno-h")
+                ann_class = gr.Dropdown(choices=read_classes(), value=(read_classes()[0] if read_classes() else None), label="Class", interactive=True)
+                x = gr.Number(label="X (left)", value=0, precision=0)
+                y = gr.Number(label="Y (top)", value=0, precision=0)
+                w = gr.Number(label="Width", value=0, precision=0)
+                h = gr.Number(label="Height", value=0, precision=0)
                 add_btn = gr.Button("💾 Save Box", variant="primary")
-                gr.Markdown("Example: X=120, Y=80, Width=180, Height=300.")
+                gr.Markdown("**Box method:** click corner 1 → click corner 2 → Save Box.")
                 delete_index = gr.Number(label="Annotation # to delete", value=1, precision=0)
                 delete_btn = gr.Button("Delete Box")
                 clear_btn = gr.Button("Clear All Boxes")
         annotations = gr.JSON(label="Saved annotations")
         ann_msg = gr.Markdown()
-        image_select.change(lambda image_id: (draw_annotations(image_id), refresh_editor(image_id)[1], refresh_editor(image_id)[2]), image_select, [annotation_canvas, editor_info, annotations])
-        add_btn.click(add_annotation, [image_select, ann_class, x, y, w, h], [annotation_canvas, ann_msg, annotations])
-        delete_btn.click(remove_annotation, [image_select, delete_index], [annotation_canvas, ann_msg, annotations])
-        clear_btn.click(clear_annotations, image_select, [annotation_canvas, ann_msg, annotations])
+        click_state = gr.State([])
+
+        def load_annotation_image(image_id):
+            return annotation_preview_with_boxes(image_id), refresh_editor(image_id)[1], refresh_editor(image_id)[2], []
+
+        image_select.change(load_annotation_image, image_select, [annotation_image, editor_info, annotations, click_state])
+        annotation_image.select(handle_annotation_click, [image_select, ann_class, click_state], [x, y, w, h, click_state, ann_msg])
+        add_btn.click(add_annotation, [image_select, ann_class, x, y, w, h], [annotation_image, ann_msg, annotations])
+        delete_btn.click(remove_annotation, [image_select, delete_index], [annotation_image, ann_msg, annotations])
+        clear_btn.click(clear_annotations, image_select, [annotation_image, ann_msg, annotations])
 
     with gr.Tab("3 · Training"):
         gr.Markdown("### Train RT-DETR")
@@ -513,9 +551,10 @@ with gr.Blocks(title="Ice Cream Dataset + Counter") as demo:
         preprocess=False,
         queue=False,
     )
+    save_class_btn.click(save_classes, class_text, [class_msg, ann_class, status])
     # The earlier placeholder event is harmlessly superseded by this real event.
 
-    demo.load(lambda: (dataset_status(), gr.update(choices=image_choices()), gr.update(choices=read_classes())),
+    demo.load(lambda: (dataset_status(), gr.update(choices=image_choices()), gr.update(choices=read_classes(), value=(read_classes()[0] if read_classes() else None))),
               None, [status, image_select, ann_class])
 
 if __name__ == "__main__":

training/classes.txt

@@ -1,8 +1,10 @@
-cornetto
-magnum
-correto
-cone
-cup
-sandwich
-stick
+Carnavalita
+Kimo-COno
+Squizz
+Oreo
+Moro
+Dulce
+KitKat
+Cadbury
+Mega
 other