Update app.py

app.py

@@ -22,7 +22,7 @@ _hf_api = HfApi(token=HF_TOKEN)
 
 # --- Main settings ---
 TARGET_PER_PERSON = 30
-CONTACT_EMAIL = "ffallah[at]asu.edu"
+CONTACT_EMAIL = "ffallah@asu.edu"
 
 # --- Paths ---
 CAPTIONS_JSON_PATH = os.environ.get("CAPTIONS_JSON_PATH", "data/captions.json")
@@ -32,6 +32,7 @@ GT_MASKED_DIR   = os.environ.get("GT_MASKED_DIR", "data/gt_masked")    # Image 1
 GT_UNMASKED_DIR = os.environ.get("GT_UNMASKED_DIR", "data/gt")        # Image 2
 SR_DIR          = os.environ.get("SR_DIR", "data/sr")               # Image 3
 ORIGINAL_DIR    = os.environ.get("ORIGINAL_DIR", "data/lr")     # Image 4
+
 # --- Results ---
 RESULTS_DIR = os.environ.get("RESULTS_DIR", "results")
 PROGRESS_PATH = os.path.join(RESULTS_DIR, "progress.json")
@@ -189,6 +190,33 @@ def append_jsonl(path: str, record: Dict[str, Any]):
         with open(path, "a", encoding="utf-8") as f:
             f.write(line + "\n")
 
+# ----------------------
+# LOGIC FOR CONVERTING SLIDERS TO RANK
+# ----------------------
+def convert_scores_to_rank(s1, s2, s3, s4) -> Dict[str, int]:
+    """
+    Takes 4 scores (0-100). Returns a dictionary:
+    {'image_1': rank, 'image_2': rank...}
+    where Rank 1 is the Highest Score.
+    """
+    # Create list of (image_key, score)
+    scores = [
+        ("image_1", s1),
+        ("image_2", s2),
+        ("image_3", s3),
+        ("image_4", s4)
+    ]
+    # Sort by score descending (High score first)
+    scores.sort(key=lambda x: x[1], reverse=True)
+    
+    ranks = {}
+    current_rank = 1
+    for img_key, score in scores:
+        ranks[img_key] = current_rank
+        current_rank += 1
+        
+    return ranks
+
 # ----------------------
 # App logic
 # ----------------------
@@ -225,7 +253,7 @@ def start_or_resume(name: str, email: str):
         )
         return (
             uid, samples, user_seen, -1,
-            None, None, None, None,None, # images and caption
+            None, None, None, None, None, # images and caption
             status,
             os.path.join(RESULTS_DIR, f"{uid}.jsonl"),
             gr.update(visible=False),
@@ -249,7 +277,6 @@ def start_or_resume(name: str, email: str):
         load_image(sample.masked_gt_path),
         load_image(sample.unmasked_gt_path),
         load_image(sample.sr_path),
-        # sample.caption,
         load_image(sample.original_path),
         f"**VLM caption:** {sample.caption}",
         status,
@@ -259,7 +286,6 @@ def start_or_resume(name: str, email: str):
         gr.update(visible=False),
     )
 
-# [CHANGE] Updated function to handle the new multiple-choice question
 def _save_record_and_progress(
     name: str,
     email: str,
@@ -267,10 +293,10 @@ def _save_record_and_progress(
     samples: List[Sample],
     user_seen: List[str],
     idx: int,
-    rank_1: str,
-    rank_2: str,
-    rank_3: str,
-    rank_4: str,
+    score_1: float,
+    score_2: float,
+    score_3: float,
+    score_4: float,
     q1_notes: str,
     q3_caption_quality: str,
 ):
@@ -280,17 +306,12 @@ def _save_record_and_progress(
     if idx is None or idx < 0 or idx >= len(samples):
         return load_progress()
 
-    # --- validate ranks ---
-    ranks = [rank_1, rank_2, rank_3, rank_4]
-    if any(r is None or r == "" for r in ranks):
-        raise gr.Error("Please rank all four images from 1 (clearest) to 4 (least clear).")
-
-    if set(ranks) != {"1", "2", "3", "4"}:
-        raise gr.Error("Each rank 1, 2, 3, 4 must be used exactly once.")
-
     if not q3_caption_quality:
         raise gr.Error("Please answer Question 2 about caption quality.")
 
+    # --- CALCULATE RANK FROM SLIDERS ---
+    rank_dict = convert_scores_to_rank(score_1, score_2, score_3, score_4)
+
     sample = samples[idx]
     progress = load_progress()
     progress.setdefault(uid, {"seen": []})
@@ -314,13 +335,15 @@ def _save_record_and_progress(
             "sr": sample.sr_path,
             "original": sample.original_path,
         },
+        # We save raw scores AND the ranking required by your analysis
+        "raw_scores": {
+            "image_1": score_1,
+            "image_2": score_2,
+            "image_3": score_3,
+            "image_4": score_4,
+        },
         "responses": {
-            "image_ranking": {
-                "image_1": int(rank_1),
-                "image_2": int(rank_2),
-                "image_3": int(rank_3),
-                "image_4": int(rank_4),
-            },
+            "image_ranking": rank_dict, # Format: {"image_1": 1, "image_2": 4, ...}
             "notes": q1_notes or "",
             "caption_quality": q3_caption_quality,
         },
@@ -336,7 +359,9 @@ def _save_record_and_progress(
     save_progress(progress)
     return progress
 
-# [CHANGE] Updated function signature
+# ----------------------
+# Buttons
+# ----------------------
 def submit_finish(
     name: str,
     email: str,
@@ -344,41 +369,35 @@ def submit_finish(
     samples: List[Sample],
     user_seen: List[str],
     idx: int,
-    rank_1: str,
-    rank_2: str,
-    rank_3: str,
-    rank_4: str,
+    s1: float, s2: float, s3: float, s4: float,
     q1_notes: str,
     q3_caption_quality: str,
 ):
     try:
         _save_record_and_progress(
             name, email, uid, samples, user_seen, idx,
-            rank_1, rank_2, rank_3, rank_4,
+            s1, s2, s3, s4,
             q1_notes, q3_caption_quality
         )
     except gr.Error:
         return (
             user_seen, idx,
-            gr.update(), gr.update(), gr.update(), gr.update(),  # images
-            gr.update(),                                         # caption_display
-            gr.update(),                                         # status
-            gr.update(),                                         # notes_q1
-            gr.update(), gr.update(), gr.update(), gr.update(),  # rank_1..4
-            gr.update(),                                         # q3
+            gr.update(), gr.update(), gr.update(), gr.update(),
+            gr.update(), gr.update(), gr.update(),
+            gr.update(), gr.update(), gr.update(), gr.update(),
+            gr.update(),
         )
 
     return (
         user_seen, idx,
-        gr.update(value=None), gr.update(value=None), gr.update(value=None), gr.update(value=None),  # images
+        gr.update(value=None), gr.update(value=None), gr.update(value=None), gr.update(value=None),
         gr.update(value=""), 
         gr.update(value="Finished!"),
         gr.update(value=""),
-        gr.update(value=None), gr.update(value=None), gr.update(value=None), gr.update(value=None),  # rank_1..4
-        gr.update(value=None),  # q3
+        gr.update(value=50), gr.update(value=50), gr.update(value=50), gr.update(value=50),
+        gr.update(value=None), 
     )
 
-# [CHANGE] Updated function signature
 def submit_next_image(
     name: str,
     email: str,
@@ -386,28 +405,23 @@ def submit_next_image(
     samples: List[Sample],
     user_seen: List[str],
     idx: int,
-    rank_1: str,
-    rank_2: str,
-    rank_3: str,
-    rank_4: str,
+    s1: float, s2: float, s3: float, s4: float,
     q1_notes: str,
     q3_caption_quality: str,
 ):
     try:
         progress = _save_record_and_progress(
             name, email, uid, samples, user_seen, idx,
-            rank_1, rank_2, rank_3, rank_4,
+            s1, s2, s3, s4,
             q1_notes, q3_caption_quality
         )
     except gr.Error:
         return (
             user_seen, idx,
-            gr.update(), gr.update(), gr.update(), gr.update(),  # images
-            gr.update(),                                         # caption_display
-            gr.update(),                                         # status
-            gr.update(),                                         # notes_q1
-            gr.update(), gr.update(), gr.update(), gr.update(),  # rank_1..4
-            gr.update(),                                         # q3
+            gr.update(), gr.update(), gr.update(), gr.update(),
+            gr.update(), gr.update(), gr.update(),
+            gr.update(), gr.update(), gr.update(), gr.update(),
+            gr.update(),
         )
 
     left_after = user_left_count(progress[uid]["seen"], samples)
@@ -420,12 +434,12 @@ def submit_next_image(
         )
         return (
             progress[uid]["seen"], idx,
-            gr.update(), gr.update(), gr.update(), gr.update(),        # images unchanged
-            gr.update(),                                               # caption_display unchanged
-            gr.update(value=status),                                   # status
-            gr.update(value=""),                                       # notes cleared
-            gr.update(value=None), gr.update(value=None), gr.update(value=None), gr.update(value=None),  # ranks cleared
-            gr.update(value=None),                                     # q3 cleared
+            gr.update(), gr.update(), gr.update(), gr.update(),        
+            gr.update(),                                               
+            gr.update(value=status),                                   
+            gr.update(value=""),                                       
+            gr.update(value=50), gr.update(value=50), gr.update(value=50), gr.update(value=50), 
+            gr.update(value=None),                                     
         )
 
     idx_next = pick_next_index(progress[uid]["seen"], samples)
@@ -443,9 +457,9 @@ def submit_next_image(
         load_image(sample_next.original_path),
         f"**VLM caption:** {sample_next.caption}",
         gr.update(value=status),
-        gr.update(value=""),                                          # notes cleared
-        gr.update(value=None), gr.update(value=None), gr.update(value=None), gr.update(value=None),  # ranks cleared
-        gr.update(value=None),                                        # q3 cleared
+        gr.update(value=""),                                          
+        gr.update(value=50), gr.update(value=50), gr.update(value=50), gr.update(value=50),
+        gr.update(value=None),                                        
     )
 
 
@@ -510,24 +524,35 @@ with gr.Blocks(title="RTS Human Evaluation", theme=gr.themes.Soft()) as demo:
 
     eval_panel = gr.Group(visible=False)
     with eval_panel:
+        
+        # --- NEW LAYOUT: SLIDERS UNDER IMAGES ---
+        gr.Markdown("### Question 1: Rate Image Clarity (Higher = Better / Clearer RTS)")
+        
         with gr.Row():
-            with gr.Column(scale=1, min_width=260):
+            with gr.Column(scale=1, min_width=200):
                 gr.Markdown("<div style='text-align:center; font-weight:600;'>Image 1</div>")
                 image_1 = gr.Image(show_label=False, interactive=False, height=256, show_download_button=False)
-            with gr.Column(scale=1, min_width=260):
+                score_1 = gr.Slider(minimum=0, maximum=100, step=1, value=50, label="Score (0-100)")
+                
+            with gr.Column(scale=1, min_width=200):
                 gr.Markdown("<div style='text-align:center; font-weight:600;'>Image 2</div>")
                 image_2 = gr.Image(show_label=False, interactive=False, height=256, show_download_button=False)
-            with gr.Column(scale=1, min_width=260):
+                score_2 = gr.Slider(minimum=0, maximum=100, step=1, value=50, label="Score (0-100)")
+                
+            with gr.Column(scale=1, min_width=200):
                 gr.Markdown("<div style='text-align:center; font-weight:600;'>Image 3</div>")
                 image_3 = gr.Image(show_label=False, interactive=False, height=256, show_download_button=False)
-            with gr.Column(scale=1, min_width=260):
+                score_3 = gr.Slider(minimum=0, maximum=100, step=1, value=50, label="Score (0-100)")
+                
+            with gr.Column(scale=1, min_width=200):
                 gr.Markdown("<div style='text-align:center; font-weight:600;'>Image 4</div>")
                 image_4 = gr.Image(show_label=False, interactive=False, height=256, show_download_button=False)
+                score_4 = gr.Slider(minimum=0, maximum=100, step=1, value=50, label="Score (0-100)")
 
         notes_q1 = gr.Textbox(
-            label="If the image contains multiple RTS, and some are clear and some are not, please add notes here.",
-            lines=3,
-            placeholder="Add any brief justification or observations…"
+            label="Notes (Optional)",
+            lines=2,
+            placeholder="If there are multiple RTS or ambiguities, please note here."
         )
         
         caption_display = gr.Markdown(label="Reference Caption")
@@ -544,23 +569,6 @@ with gr.Blocks(title="RTS Human Evaluation", theme=gr.themes.Soft()) as demo:
         )
 
         with gr.Row():
-
-            rank_1 = gr.Dropdown(
-                label="Question 1: Rank for Image 1 (1 = clearest RTS in the bounding box, 4 = least clear)",
-                choices=["1", "2", "3", "4"]
-            )
-            rank_2 = gr.Dropdown(
-                label="Rank for Image 2 (1 = clearest, 4 = least clear)",
-                choices=["1", "2", "3", "4"]
-            )
-            rank_3 = gr.Dropdown(
-                label="Rank for Image 3 (1 = clearest, 4 = least clear)",
-                choices=["1", "2", "3", "4"]
-            )
-            rank_4 = gr.Dropdown(
-                label="Rank for Image 4 (1 = clearest, 4 = least clear)",
-                choices=["1", "2", "3", "4"]
-            )
             submit_next_btn = gr.Button("Submit & Next Image", variant="primary")
             submit_finish_btn = gr.Button("Submit & Finish", variant="secondary")
 
@@ -572,14 +580,6 @@ with gr.Blocks(title="RTS Human Evaluation", theme=gr.themes.Soft()) as demo:
 
     # --- Wiring ---
 
-    # def toggle_q2_visibility(q1_choice):
-    #     if q1_choice == "No":
-    #         return gr.update(visible=True)
-    #     else:
-    #         return gr.update(visible=True, value=None)
-
-    # q1.change(toggle_q2_visibility, inputs=q1, outputs=q2)
-
     start_event = start_btn.click(
         start_or_resume,
         inputs=[name, email],
@@ -595,11 +595,11 @@ with gr.Blocks(title="RTS Human Evaluation", theme=gr.themes.Soft()) as demo:
     finish_event = submit_finish_btn.click(
         submit_finish,
         inputs=[name, email, state_uid, state_samples, state_seen, state_idx,
-                rank_1, rank_2, rank_3, rank_4, notes_q1, q3],
+                score_1, score_2, score_3, score_4, notes_q1, q3],
         outputs=[state_seen, state_idx,
                 image_1, image_2, image_3, image_4,
                 caption_display, status, notes_q1,
-                rank_1, rank_2, rank_3, rank_4, q3],
+                score_1, score_2, score_3, score_4, q3],
     )
     finish_event.then(
         to_thanks,
@@ -610,11 +610,11 @@ with gr.Blocks(title="RTS Human Evaluation", theme=gr.themes.Soft()) as demo:
     nextimg_event = submit_next_btn.click(
         submit_next_image,
         inputs=[name, email, state_uid, state_samples, state_seen, state_idx,
-                rank_1, rank_2, rank_3, rank_4, notes_q1, q3],
+                score_1, score_2, score_3, score_4, notes_q1, q3],
         outputs=[state_seen, state_idx,
                 image_1, image_2, image_3, image_4,
                 caption_display, status, notes_q1,
-                rank_1, rank_2, rank_3, rank_4, q3],
+                score_1, score_2, score_3, score_4, q3],
     )
     nextimg_event.then(
         maybe_show_thanks,