From 9fe61de82c1a555d4b0e4aa841f0361f64ae440c Mon Sep 17 00:00:00 2001
From: JianHe0628 <jl02958@surrey.ac.uk>
Date: Wed, 29 Jan 2025 17:42:07 +0000
Subject: [PATCH] Updating code for CSLdaily
---
Extract_HaMeR_Single.py | 13 +++++++++----
Inject_Json.py | 16 ++++++++--------
2 files changed, 17 insertions(+), 12 deletions(-)
diff --git a/Extract_HaMeR_Single.py b/Extract_HaMeR_Single.py
index a10f7c8..95f7ca8 100644
--- a/Extract_HaMeR_Single.py
+++ b/Extract_HaMeR_Single.py
@@ -118,7 +118,7 @@ def main(args, model, renderer, device):
img_paths = [img for end in args.file_type for img in image_folder.glob(end)]
#VideoFile Specific
- img_paths = sorted(img_paths,key = lambda x: int(os.path.basename(x).removesuffix('.png').removeprefix('Frame')))
+ img_paths = sorted(img_paths,key = lambda x: int(os.path.basename(x).removesuffix('.png').removesuffix('.jpg').removeprefix('Frame')))
# #For Phoenix
# img_paths = sorted(img_paths,key = lambda x: int(os.path.basename(x).removeprefix('images').removesuffix('.png')))
@@ -144,6 +144,9 @@ def main(args, model, renderer, device):
else:
print("No humans detected in the image")
return
+ print(f"Using detected bbox, {pred_bboxes[0]}")
+
+
else:
# Hardcoded bbox. This assumes person is in the center and that there is always one person in the image
if args.custom_bbox == '':
@@ -156,7 +159,7 @@ def main(args, model, renderer, device):
# Force confidence to be 0.99 that human is present
pred_scores = np.array([0.99])
-
+ # show_bbox(img_cv2, pred_bboxes)
# keypoint detector
cpm = ViTPoseModel(device)
print('Loading ViTPose Model')
@@ -301,8 +304,10 @@ if __name__ == '__main__':
args = parser.parse_args()
args.bbox = 'False'
- args.custom_bbox = '70,0,550,360'
- args.out_folder = os.path.join(args.out_folder, os.path.dirname(args.vid).removeprefix('/vol/vssp/LF_datasets/multiview/bslcp/BSLCP/'))
+ args.custom_bbox = '0,60,512,512'
+ # args.img_folder = '/vol/vssp/datasets/singlevideo/csldaily/sentences/frames_512x512/S000000_P0000_T00'
+ # args.out_folder = '/vol/research/signVision/Projects/BSLboundaries/Testing'
+ args.out_folder = os.path.join(args.out_folder, os.path.basename(args.img_folder))
print(f"Output folder: {args.out_folder}")
model, model_cfg = load_hamer(args.checkpoint, load_mesh=False) # False sets model to not produce vertice on inference
diff --git a/Inject_Json.py b/Inject_Json.py
index 27a55aa..fad14e1 100644
--- a/Inject_Json.py
+++ b/Inject_Json.py
@@ -138,7 +138,7 @@ def main(args, model, device):
img_paths = [img for end in args.file_type for img in image_folder.glob(end)]
# For General Video
- img_paths = sorted(img_paths,key = lambda x: int(os.path.basename(x).removesuffix('.png').removeprefix('Frame')))
+ img_paths = sorted(img_paths,key = lambda x: int(os.path.basename(x).removesuffix('.png').removesuffix('.jpg').removeprefix('Frame')))
# #For Phoenix
# img_paths = sorted(img_paths,key = lambda x: int(os.path.basename(x).removeprefix('images').removesuffix('.png')))
assert len(img_paths) == len(hand_pose_list), f"Number of images {len(img_paths)} and lmdb files {len(hand_pose_list)} do not match"
@@ -172,6 +172,7 @@ def main(args, model, device):
else:
print("No humans detected in the first frame. Exiting...")
return
+
else:
# Hardcoded bbox. This assumes person is in the center and that there is always one person in the image
if args.custom_bbox == '':
@@ -351,15 +352,14 @@ if __name__ == '__main__':
args = parser.parse_args()
- Filename = 'BUSINESS_GROUP_AND_SET-UP_SPIN-OFF_PARTNER-RELATIONSHIP'
+ args.img_folder = '/vol/vssp/datasets/singlevideo/csldaily/sentences/frames_512x512/S000000_P0000_T00'
- args.vid = f'/vol/vssp/LF_datasets/multiview/bslcp/BSLCP/Videos/Narratives/Belfast/17+18/BF17n.mov'
- args.lmdb_input = f'/mnt/fast/nobackup/scratch4weeks/jl02958/BSLCorpus/Videos/Narratives/Belfast/17+18/BF17n.lmdb'
- args.out_folder = f'/vol/research/signVision/Projects/BSLboundaries/Test/BF17n'
- args.frame_range = '30,400'
+ args.lmdb_input = f'/vol/research/signVision/Projects/BSLboundaries/Test/S000000_P0000_T00/S000000_P0000_T00.lmdb'
+ args.out_folder = f'/vol/research/signVision/Projects/BSLboundaries/Test/Check_S000000_P0000_T00'
+ args.frame_range = '0,50'
args.bbox = False
- args.custom_bbox = '70,0,550,360'
- args.save_mesh = False
+ args.custom_bbox = '0,60,512,512'
+ args.save_mesh = True
# Download and load checkpoints
# download_models(CACHE_DIR_HAMER)
--
GitLab