From d5919041fc0d2e5a53dfb012b3b526f6113cf0c8 Mon Sep 17 00:00:00 2001
From: Han Wang <109300432+Hon-Wong@users.noreply.github.com>
Date: Fri, 25 Oct 2024 16:40:59 +0800
Subject: [PATCH] Update merge_result.py

---
 eval/merge_result.py | 110 ++++++++++++++++++++++---------------------
 1 file changed, 56 insertions(+), 54 deletions(-)

diff --git a/eval/merge_result.py b/eval/merge_result.py
index 5059ba1..fb85cd4 100644
--- a/eval/merge_result.py
+++ b/eval/merge_result.py
@@ -35,59 +35,61 @@ def b2str(b):
     return "[" + ",".join([str(int(pos)) for pos in b]) + "]"
 
 results = {}
-for jsonfile in args.files_to_merge:
-    print(jsonfile)
-    f = open(jsonfile, "r").read()
-    # flat_outputs = json.loads(f.read())
-    lines = [l + "}" for l in f.split("}\n")]
-    start_frame_idx = 0
-    is_first_line = True
-    for line in lines:
-        try:
-            line = json.loads(line)
-        except:
-            if len(line) < 4:
-                continue
-            # print(line)
-        _id = line["id"]
-        seq_id, clip_id = _id.split("|")
-        seq_id = seq_id
-        if seq_id not in results: 
-            is_first_line = True
-            start_frame_idx = 0
-        else:
-            is_first_line = False
-        clip_id = int(clip_id)
-        image_size = line["image_size"]
-        predict = line["predict"]
-        gt = line["gt"]
-        pred_bb = parse_box_from_raw_text(predict)
-        anno_bb = parse_box_from_raw_text(gt)
-        if len(pred_bb) > len(anno_bb):
-            pred_bb = pred_bb[:len(anno_bb)]
-        elif len(pred_bb) < len(anno_bb):
-            pad_len = len(anno_bb) - len(pred_bb)
-            for i in range(pad_len):
-                pred_bb.append([0., 0., 0., 0.])
-        clip_frame_count = len(pred_bb) if is_first_line else len(pred_bb) - 1
+jsonfile = args.files_to_merge
+print(jsonfile)
+f = open(jsonfile, "r").read()
+# flat_outputs = json.loads(f.read())
+lines = [l + "}" for l in f.split("}\n")]
+start_frame_idx = 0
+is_first_line = True
+for line in lines:
+    try:
+        line = json.loads(line)
+    except:
+        if len(line) < 4:
+            continue
+        # print(line)
+    _id = line["id"]
+    seq_id, clip_id = _id.split("|")
+    seq_id = seq_id
+    if seq_id not in results: 
+        is_first_line = True
+        start_frame_idx = 0
+    else:
+        is_first_line = False
+    clip_id = int(clip_id)
+    image_size = line["image_size"]
+    predict = line["predict"]
+    gt = line["gt"]
+    pred_bb = parse_box_from_raw_text(predict)
+    anno_bb = parse_box_from_raw_text(gt)
+    if len(pred_bb) > len(anno_bb):
+        pred_bb = pred_bb[:len(anno_bb)]
+    elif len(pred_bb) < len(anno_bb):
+        pad_len = len(anno_bb) - len(pred_bb)
+        for i in range(pad_len):
+            pred_bb.append([0., 0., 0., 0.])
+    clip_frame_count = len(pred_bb) if is_first_line else len(pred_bb) - 1
 
-        pred_bb = pred_bb[-clip_frame_count:]
-        anno_bb = anno_bb[-clip_frame_count:]
- 
-        if seq_id not in results: 
-            results[seq_id] = {}
-            results[seq_id]["predict"] = ",".join(f"Frame {start_frame_idx + i + 1}: {b2str(pred_bb[i])}" for i in range(0, clip_frame_count))
-            results[seq_id]["gt"] =  ",".join(f"Frame {start_frame_idx + i + 1}: {b2str(anno_bb[i])}" for i in range(0, clip_frame_count))
-            results[seq_id]["image_size"] = image_size
-            results[seq_id]["source"] = "unknown"
-            results[seq_id]["vid"] = seq_id
-            results[seq_id]["id"] = seq_id
-            clip_count = 1
-        else:
-            clip_count += 1
-            if clip_count < 1000000000:  # see if metrics go lower when processing longer frames, set to a large value to evaluate the whole video
-                results[seq_id]["predict"] += ",".join(f"Frame {start_frame_idx + i + 1}: {b2str(pred_bb[i])}" for i in range(0, clip_frame_count))
-                results[seq_id]["gt"] += ",".join(f"Frame {start_frame_idx + i + 1}: {b2str(anno_bb[i])}" for i in range(0, clip_frame_count))
-        start_frame_idx += clip_frame_count
+    pred_bb = pred_bb[-clip_frame_count:]
+    anno_bb = anno_bb[-clip_frame_count:]
+
+    if seq_id not in results: 
+        results[seq_id] = {}
+        results[seq_id]["predict"] = ",".join(f"Frame {start_frame_idx + i + 1}: {b2str(pred_bb[i])}" for i in range(0, clip_frame_count))
+        results[seq_id]["gt"] =  ",".join(f"Frame {start_frame_idx + i + 1}: {b2str(anno_bb[i])}" for i in range(0, clip_frame_count))
+        results[seq_id]["image_size"] = image_size
+        results[seq_id]["source"] = "unknown"
+        results[seq_id]["vid"] = seq_id
+        results[seq_id]["id"] = seq_id
+        clip_count = 1
+    else:
+        clip_count += 1
+        if clip_count < 1000000000:  # see if metrics go lower when processing longer frames, set to a large value to evaluate the whole video
+            results[seq_id]["predict"] += ",".join(f"Frame {start_frame_idx + i + 1}: {b2str(pred_bb[i])}" for i in range(0, clip_frame_count))
+            results[seq_id]["gt"] += ",".join(f"Frame {start_frame_idx + i + 1}: {b2str(anno_bb[i])}" for i in range(0, clip_frame_count))
+    start_frame_idx += clip_frame_count
 
-json.dump(list(results.values()), open(args.output_file, "w"), indent=4, ensure_ascii=False)
+with open(args.output_file, "w") as f:
+    for line in list(results.values()):
+        f.write(json.dumps(line, ensure_ascii=False) + '\n')