Initial commit

app.py

@@ -9,18 +9,21 @@ from concurrent.futures import ThreadPoolExecutor
 from datetime import datetime
 from filelock import FileLock 
 
-# --- CONFIGURATION ---
+if os.path.exists("/data"):
+    DATA_DIR = "/data"
+else:
+    DATA_DIR = "."
+
 URL_FILE = "urls.txt"
-LABEL_FILE = "annotations.csv"
-VERIFY_FILE = "verifications.csv"
-SKIP_FILE = "skipped.csv"
-LOCK_FILE = "data.lock" 
+LABEL_FILE = os.path.join(DATA_DIR, "annotations.csv")
+VERIFY_FILE = os.path.join(DATA_DIR, "verifications.csv")
+SKIP_FILE = os.path.join(DATA_DIR, "skipped.csv")
+LOCK_FILE = os.path.join(DATA_DIR, "data.lock")
 
 MAX_IMAGES = 6
-THUMB_SIZE = (350, 350) # Reducing size drastically speeds up rendering
+THUMB_SIZE = (350, 350)
 ROOM_CLASSES = ["living_room", "bedroom", "kitchen", "bathroom", "dining_room", "outdoor", "other"]
 
-# --- INIT ---
 def init_files():
     for f in [LABEL_FILE, VERIFY_FILE, SKIP_FILE]:
         if not os.path.exists(f):
@@ -33,29 +36,20 @@ def init_files():
 
 init_files()
 
-# --- FAST IMAGE ENGINE ---
-
 def get_image_optimized(url):
-    """
-    Downloads high-res image, resizes to thumbnail, returns PIL Object.
-    If fails, returns the URL string (fallback).
-    """
     if not url: return None
     try:
-        # Pretend to be Chrome to avoid blocking
         headers = {
             'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'
         }
-        # 3 second timeout prevents freezing
         response = requests.get(url, headers=headers, timeout=3)
         if response.status_code == 200:
             img = Image.open(BytesIO(response.content))
-            img.thumbnail(THUMB_SIZE, Image.Resampling.LANCZOS) # Resize
+            img.thumbnail(THUMB_SIZE, Image.Resampling.LANCZOS)
             return img
     except:
         pass
     
-    # Fallback: If server blocks Python, return URL so browser tries
     return url 
 
 def get_ordered_groups():
@@ -103,13 +97,10 @@ def get_stats_text():
     except: l = 0
     return f"**Total Properties:** {len(all_gids)} | **Labeled:** {l}"
 
-# --- APP LOGIC ---
-
 def render_workspace(mode, history, specific_index=None, move_back=False):
     all_groups = get_ordered_groups()
     total_groups = len(all_groups)
     
-    # Load Status (Cached reading could improve this further, but pandas is fast enough for <10k rows)
     try: l_done = set(pd.read_csv(LABEL_FILE)['group_id'].unique())
     except: l_done = set()
     try: v_done = set(pd.read_csv(VERIFY_FILE)['group_id'].unique())
@@ -120,7 +111,6 @@ def render_workspace(mode, history, specific_index=None, move_back=False):
     target_gid = None
     target_idx = -1
 
-    # 1. NAVIGATION
     if specific_index is not None:
         if 0 <= specific_index < total_groups:
             target_gid = all_groups[specific_index]
@@ -135,7 +125,6 @@ def render_workspace(mode, history, specific_index=None, move_back=False):
         except: target_idx = 0
         
     else:
-        # Auto-Next
         found = False
         for i, gid in enumerate(all_groups):
             if gid in s_done: continue
@@ -152,7 +141,6 @@ def render_workspace(mode, history, specific_index=None, move_back=False):
         if not found:
              return {screen_menu: gr.update(visible=True), screen_work: gr.update(visible=False), log_box: "No more tasks found."}
 
-    # 2. PREPARE DATA
     urls = get_group_urls(target_gid)
     
     if not history or history[-1] != target_gid:
@@ -161,8 +149,6 @@ def render_workspace(mode, history, specific_index=None, move_back=False):
     saved_vals = get_saved_values(target_gid, mode)
     r1_vals = get_saved_values(target_gid, "label") if mode == "verify" else {}
 
-    # 3. PARALLEL IMAGE PROCESSING (THE SPEED FIX)
-    # We download and resize all 6 images at the same time
     processed_images = [None] * MAX_IMAGES
     with ThreadPoolExecutor(max_workers=MAX_IMAGES) as executor:
         futures = {executor.submit(get_image_optimized, u): i for i, u in enumerate(urls)}
@@ -175,14 +161,13 @@ def render_workspace(mode, history, specific_index=None, move_back=False):
         screen_menu: gr.update(visible=False),
         screen_work: gr.update(visible=True),
         header_md: header,
-        state_urls: urls, # Keep raw URLs for saving
+        state_urls: urls,
         state_hist: history,
         state_idx: target_idx,
         top_stats: get_stats_text(),
         log_box: f"Loaded group {target_gid}"
     }
 
-    # 4. UPDATE GRID
     for i in range(MAX_IMAGES):
         img_c = img_objs[i]
         base = i * 4
@@ -190,7 +175,7 @@ def render_workspace(mode, history, specific_index=None, move_back=False):
         
         if i < len(urls):
             u = urls[i]
-            img_data = processed_images[i] # This is now a small PIL Image or a URL string
+            img_data = processed_images[i]
             
             updates[img_c] = gr.update(value=img_data, visible=True)
             
@@ -262,8 +247,6 @@ def refresh_cat():
         data.append([i+1, s, gid])
     return pd.DataFrame(data, columns=["#", "Status", "ID"])
 
-# --- UI ---
-
 with gr.Blocks(title="Fast Labeler") as demo:
     
     state_mode = gr.State("label")
@@ -314,7 +297,6 @@ with gr.Blocks(title="Fast Labeler") as demo:
             df_cat = gr.Dataframe(interactive=False)
             b_ref_cat = gr.Button("Refresh")
 
-    # WIRING
     ALL_IO = [screen_menu, screen_work, header_md, state_urls, state_hist, state_idx, top_stats, log_box] + img_objs + input_objs
     
     b_start_l.click(lambda: "label", None, state_mode).then(render_workspace, [state_mode, state_hist], ALL_IO)
@@ -332,4 +314,4 @@ with gr.Blocks(title="Fast Labeler") as demo:
     b_ref_cat.click(refresh_cat, None, df_cat)
     demo.load(refresh_cat, None, df_cat).then(get_stats_text, None, top_stats)
 
-demo.queue().launch(theme=gr.themes.Soft())
+demo.queue().launch(theme=gr.themes.Soft())