lzy/sci-gui-agent-benchmark
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: fix multiapp tasks (#229)

Browse Source 
Co-authored-by: adlsdztony <zzl0712@connect.hku.hk>
This commit is contained in:
XXZ
2025-07-03 21:53:58 +08:00
committed by GitHub
parent 7b2120c843
commit ac24ccce99
12 changed files with 57 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
desktop_env/evaluators/metrics/gimp.py
View File

@@ -193,7 +193,7 @@ def structure_check_by_mse(img1, img2, threshold=0.03):
(np.array(img1, dtype=np.float32) / 255
- np.array(img2, dtype=np.float32) / 255) ** 2)
structure_same = True if mse < threshold else False
print("MSE: ", mse)
print(f"MSE: {mse}, threshold: {threshold}")
return structure_same
@@ -204,7 +204,7 @@ def structure_check_by_ssim(img1, img2, threshold=0.9):
return similarity >= threshold
def check_brightness_decrease_and_structure_sim(src_path, tgt_path):
def check_brightness_decrease_and_structure_sim(src_path, tgt_path, threshold=0.03):
"""
Check the brightness of src is lower than tgt and the structures are similar
gimp:7a4deb26-d57d-4ea9-9a73-630f66a7b568
@@ -219,13 +219,15 @@ def check_brightness_decrease_and_structure_sim(src_path, tgt_path):
brightness_src = calculate_brightness(img_src)
brightness_tgt = calculate_brightness(img_tgt)
brightness_reduced = brightness_tgt > brightness_src
# print(f"Brightness src: {brightness_src}, tgt: {brightness_tgt}, reduced: {brightness_reduced}")
# Normalize and compare images
target_brightness = 128
img_src_normalized = normalize_brightness(img_src, target_brightness)
img_tgt_normalized = normalize_brightness(img_tgt, target_brightness)
structure_same = structure_check_by_mse(img_src_normalized, img_tgt_normalized)
structure_same = structure_check_by_mse(img_src_normalized, img_tgt_normalized, threshold=threshold)
if brightness_reduced and structure_same:
return 1.
else:

6
desktop_env/evaluators/metrics/others.py
View File

@@ -63,11 +63,13 @@ def compare_epub(result: str, expected: str) -> float:
result_files: List[str] = process_epub(result)
expected_files: List[str] = process_epub(expected)
metric: float = 1.
metric: float = 0.
for f1, f2 in zip(result_files, expected_files):
current_metric: float = diff_text_file(f1, f2)
logger.debug("%s vs %s: %f", f1, f2, current_metric)
metric *= current_metric
metric += current_metric
if len(result_files) > 0:
metric /= len(result_files)
return metric