File size: 27,022 Bytes
26e5c98
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ea97ae7
26e5c98
ea97ae7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26e5c98
 
 
 
 
 
5fb16c4
 
26e5c98
 
 
ea97ae7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26e5c98
 
ea97ae7
26e5c98
ea97ae7
26e5c98
ea97ae7
 
26e5c98
ea97ae7
26e5c98
ea97ae7
 
 
 
26e5c98
ea97ae7
 
 
 
26e5c98
ea97ae7
 
 
 
 
26e5c98
ea97ae7
 
 
 
26e5c98
ea97ae7
 
 
 
26e5c98
ea97ae7
 
26e5c98
ea97ae7
26e5c98
ea97ae7
 
26e5c98
ea97ae7
 
 
 
 
 
26e5c98
ea97ae7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26e5c98
ea97ae7
 
 
 
26e5c98
ea97ae7
 
 
 
 
 
 
26e5c98
ea97ae7
26e5c98
 
ea97ae7
26e5c98
 
ea97ae7
26e5c98
 
 
ea97ae7
26e5c98
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ea97ae7
26e5c98
 
ea97ae7
26e5c98
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ea97ae7
 
 
 
 
 
 
 
 
 
 
 
 
26e5c98
ea97ae7
 
 
26e5c98
ea97ae7
 
 
 
 
 
 
 
26e5c98
ea97ae7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
26e5c98
ea97ae7
 
 
 
26e5c98
 
ea97ae7
26e5c98
ea97ae7
 
 
 
26e5c98
ea97ae7
26e5c98
 
 
ea97ae7
 
 
 
 
 
 
 
 
 
 
 
26e5c98
 
ea97ae7
 
26e5c98
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ea97ae7
 
 
26e5c98
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ea97ae7
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
# Copyright 2024-2025 The Alibaba Wan Team Authors. All rights reserved.
import os
import numpy as np
import shutil
import torch
from diffusers import FluxKontextPipeline
import cv2
from loguru import logger
from PIL import Image
try:
    import moviepy.editor as mpy
except:
    import moviepy as mpy

from decord import VideoReader
from pose2d import Pose2d
from pose2d_utils import AAPoseMeta
from utils import resize_by_area, get_frame_indices, padding_resize, get_face_bboxes, get_aug_mask, get_mask_body_img
from human_visualization import draw_aapose_by_meta_new
from retarget_pose import get_retarget_pose
from sam2.build_sam import build_sam2, build_sam2_video_predictor

def get_frames(video_path, resolution_area, fps=30):
    video_reader = VideoReader(video_path)
    frame_num = len(video_reader)
    
    video_fps = video_reader.get_avg_fps()


    # TODO: Maybe we can switch to PyAV later, which can get accurate frame num
    duration = video_reader.get_frame_timestamp(-1)[-1]
  
    expected_frame_num = int(duration * video_fps + 0.5) 
    ratio = abs((frame_num - expected_frame_num)/frame_num)         
    if ratio > 0.1:
        print("Warning: The difference between the actual number of frames and the expected number of frames is two large")
        frame_num = expected_frame_num

    if fps == -1:
        fps = video_fps

    target_num = int(frame_num / video_fps * fps)
    idxs = get_frame_indices(frame_num, video_fps, target_num, fps)
    frames = video_reader.get_batch(idxs).asnumpy()
    frames = [resize_by_area(frame, resolution_area[0] * resolution_area[1], divisor=16) for frame in frames]

    return frames

def quantize_mask_blocky(mask, block_w=16, block_h=16, occupancy=0.15):
    """
    Convert a binary mask to a blocky (quantized) mask.
    - block_w, block_h: target block size in pixels
    - occupancy: fraction [0..1] of foreground within a block to turn it on
    """
    m = (mask > 0).astype(np.uint8)
    H, W = m.shape[:2]

    # compute “block grid” size
    grid_w = max(1, int(np.ceil(W / block_w)))
    grid_h = max(1, int(np.ceil(H / block_h)))

    # downsample to grid using area interpolation (captures occupancy)
    small = cv2.resize(m, (grid_w, grid_h), interpolation=cv2.INTER_AREA)

    # threshold by occupancy (values now in [0,1] if source was 0/1)
    small_q = (small >= occupancy).astype(np.uint8)

    # upsample back with nearest (keeps sharp blocks)
    blocky = cv2.resize(small_q, (W, H), interpolation=cv2.INTER_NEAREST)
    return blocky

class ProcessPipeline():
    def __init__(self, det_checkpoint_path, pose2d_checkpoint_path, sam_checkpoint_path, flux_kontext_path):
        self.pose2d = Pose2d(checkpoint=pose2d_checkpoint_path, detector_checkpoint=det_checkpoint_path)

        if sam_checkpoint_path is not None:
            model_cfg = sam_checkpoint_path[1]
            self.predictor = build_sam2_video_predictor(model_cfg, sam_checkpoint_path[0], device="cuda")    
        if flux_kontext_path is not None:
            self.flux_kontext = FluxKontextPipeline.from_pretrained(flux_kontext_path, torch_dtype=torch.bfloat16).to("cuda")


    def __call__(self, video_path, 
                 refer_image_path, 
                 output_path, 
                 resolution_area=[1280, 720], 
                 fps=30, iterations=3, 
                 k=7, 
                 w_len=1, 
                 h_len=1, 
                 retarget_flag=False, 
                 use_flux=False, 
                 replace_flag=False,
                 pts_by_frame=None,
                 lbs_by_frame=None):
        
        if replace_flag:

            frames = get_frames(video_path, resolution_area, fps)

            height, width = frames[0].shape[:2]

            if not pts_by_frame and not lbs_by_frame:
                ############################################################################

                tpl_pose_metas = self.pose2d(frames)

                face_images = []
                for idx, meta in enumerate(tpl_pose_metas):
                    face_bbox_for_image = get_face_bboxes(meta['keypoints_face'][:, :2], scale=1.3,
                                                        image_shape=(frames[0].shape[0], frames[0].shape[1]))

                    x1, x2, y1, y2 = face_bbox_for_image
                    face_image = frames[idx][y1:y2, x1:x2]
                    face_image = cv2.resize(face_image, (512, 512))
                    face_images.append(face_image)

                logger.info(f"Processing reference image: {refer_image_path}")
                refer_img = cv2.imread(refer_image_path)
                src_ref_path = os.path.join(output_path, 'src_ref.png')
                shutil.copy(refer_image_path, src_ref_path)
                refer_img = refer_img[..., ::-1]

                refer_img = padding_resize(refer_img, height, width)
                logger.info(f"Processing template video: {video_path}")
                tpl_retarget_pose_metas = [AAPoseMeta.from_humanapi_meta(meta) for meta in tpl_pose_metas]
                cond_images = []

                for idx, meta in enumerate(tpl_retarget_pose_metas):
                    canvas = np.zeros_like(refer_img)
                    conditioning_image = draw_aapose_by_meta_new(canvas, meta)
                    cond_images.append(conditioning_image)

                ############################################################################
                ############################################################################

                masks = self.get_mask_from_face_bbox(frames, 400, tpl_pose_metas)

                bg_images = []
                aug_masks = []

                for frame, mask in zip(frames, masks):
                    if iterations > 0:
                        _, each_mask = get_mask_body_img(frame, mask, iterations=iterations, k=k)
                        each_aug_mask = get_aug_mask(each_mask, w_len=w_len, h_len=h_len)
                    else:
                        each_aug_mask = mask

                    each_bg_image = frame * (1 - each_aug_mask[:, :, None])
                    bg_images.append(each_bg_image)
                    aug_masks.append(each_aug_mask)

                ############################################################################
            else:
                ############################################################################

                masks = self.get_mask_from_face_bbox_v2(frames, pts_by_frame=pts_by_frame, lbs_by_frame=lbs_by_frame)

                bg_images = []
                aug_masks = []

                for frame, mask in zip(frames, masks):
                    if iterations > 0:
                        _, each_mask = get_mask_body_img(frame, mask, iterations=iterations, k=k)
                        # each_aug_mask = get_aug_mask(each_mask, w_len=w_len, h_len=h_len)
                        each_aug_mask = quantize_mask_blocky(each_mask, block_w=16, block_h=16, occupancy=0.15)
                        # each_aug_mask = each_mask
                    else:
                        each_aug_mask = mask

                    each_bg_image = frame * (1 - each_aug_mask[:, :, None])
                    bg_images.append(each_bg_image)
                    aug_masks.append(each_aug_mask)

                ############################################################################    
                ############################################################################
                tpl_pose_metas = self.pose2d(
                    frames,
                    bbx=masks,          # your per-frame masks list/array
                )


                face_images = []
                for idx, meta in enumerate(tpl_pose_metas):
                    face_bbox_for_image = get_face_bboxes(meta['keypoints_face'][:, :2], scale=1.3,
                                                        image_shape=(frames[0].shape[0], frames[0].shape[1]))

                    x1, x2, y1, y2 = face_bbox_for_image
                    face_image = frames[idx][y1:y2, x1:x2]
                    face_image = cv2.resize(face_image, (512, 512))
                    face_images.append(face_image)

                logger.info(f"Processing reference image: {refer_image_path}")
                refer_img = cv2.imread(refer_image_path)
                src_ref_path = os.path.join(output_path, 'src_ref.png')
                shutil.copy(refer_image_path, src_ref_path)
                refer_img = refer_img[..., ::-1]

                refer_img = padding_resize(refer_img, height, width)
                logger.info(f"Processing template video: {video_path}")
                tpl_retarget_pose_metas = [AAPoseMeta.from_humanapi_meta(meta) for meta in tpl_pose_metas]
                cond_images = []

                for idx, meta in enumerate(tpl_retarget_pose_metas):
                    canvas = np.zeros_like(refer_img)
                    conditioning_image = draw_aapose_by_meta_new(canvas, meta)
                    cond_images.append(conditioning_image)

                ############################################################################           
            
            src_face_path = os.path.join(output_path, 'src_face.mp4')
            mpy.ImageSequenceClip(face_images, fps=fps).write_videofile(src_face_path, logger=None)

            src_pose_path = os.path.join(output_path, 'src_pose.mp4')
            mpy.ImageSequenceClip(cond_images, fps=fps).write_videofile(src_pose_path, logger=None)

            src_bg_path = os.path.join(output_path, 'src_bg.mp4')
            mpy.ImageSequenceClip(bg_images, fps=fps).write_videofile(src_bg_path, logger=None)

            aug_masks_new = [np.stack([mask * 255, mask * 255, mask * 255], axis=2) for mask in aug_masks]
            src_mask_path = os.path.join(output_path, 'src_mask.mp4')
            mpy.ImageSequenceClip(aug_masks_new, fps=fps).write_videofile(src_mask_path, logger=None)
            return True
        else:
            logger.info(f"Processing reference image: {refer_image_path}")
            refer_img = cv2.imread(refer_image_path)
            src_ref_path = os.path.join(output_path, 'src_ref.png')
            shutil.copy(refer_image_path, src_ref_path)
            refer_img = refer_img[..., ::-1]
            
            refer_img = resize_by_area(refer_img, resolution_area[0] * resolution_area[1], divisor=16)
            
            refer_pose_meta = self.pose2d([refer_img])[0]


            logger.info(f"Processing template video: {video_path}")
            video_reader = VideoReader(video_path)
            frame_num = len(video_reader)

            video_fps = video_reader.get_avg_fps()

            # TODO: Maybe we can switch to PyAV later, which can get accurate frame num
            duration = video_reader.get_frame_timestamp(-1)[-1]      
            expected_frame_num = int(duration * video_fps + 0.5) 
            ratio = abs((frame_num - expected_frame_num)/frame_num)         
            if ratio > 0.1:
                print("Warning: The difference between the actual number of frames and the expected number of frames is two large")
                frame_num = expected_frame_num

            if fps == -1:
                fps = video_fps
                
            target_num = int(frame_num / video_fps * fps)
            idxs = get_frame_indices(frame_num, video_fps, target_num, fps)
            frames = video_reader.get_batch(idxs).asnumpy()

            logger.info(f"Processing pose meta")

            tpl_pose_meta0 = self.pose2d(frames[:1])[0]
            tpl_pose_metas = self.pose2d(frames)

            face_images = []
            for idx, meta in enumerate(tpl_pose_metas):
                face_bbox_for_image = get_face_bboxes(meta['keypoints_face'][:, :2], scale=1.3,
                                                    image_shape=(frames[0].shape[0], frames[0].shape[1]))

                x1, x2, y1, y2 = face_bbox_for_image
                face_image = frames[idx][y1:y2, x1:x2]
                face_image = cv2.resize(face_image, (512, 512))
                face_images.append(face_image)

            if retarget_flag:
                if use_flux:
                    tpl_prompt, refer_prompt = self.get_editing_prompts(tpl_pose_metas, refer_pose_meta)
                    refer_input = Image.fromarray(refer_img)
                    refer_edit = self.flux_kontext(
                            image=refer_input,
                            height=refer_img.shape[0],
                            width=refer_img.shape[1],
                            prompt=refer_prompt,
                            guidance_scale=2.5,
                            num_inference_steps=28,
                        ).images[0]
                    
                    refer_edit = Image.fromarray(padding_resize(np.array(refer_edit), refer_img.shape[0], refer_img.shape[1]))
                    refer_edit_path = os.path.join(output_path, 'refer_edit.png')
                    refer_edit.save(refer_edit_path)
                    refer_edit_pose_meta = self.pose2d([np.array(refer_edit)])[0]

                    tpl_img = frames[1]
                    tpl_input = Image.fromarray(tpl_img)
                    
                    tpl_edit = self.flux_kontext(
                            image=tpl_input,
                            height=tpl_img.shape[0],
                            width=tpl_img.shape[1],
                            prompt=tpl_prompt,
                            guidance_scale=2.5,
                            num_inference_steps=28,
                        ).images[0]
                    
                    tpl_edit = Image.fromarray(padding_resize(np.array(tpl_edit), tpl_img.shape[0], tpl_img.shape[1]))
                    tpl_edit_path = os.path.join(output_path, 'tpl_edit.png')
                    tpl_edit.save(tpl_edit_path)
                    tpl_edit_pose_meta0 = self.pose2d([np.array(tpl_edit)])[0]
                    tpl_retarget_pose_metas = get_retarget_pose(tpl_pose_meta0, refer_pose_meta, tpl_pose_metas, tpl_edit_pose_meta0, refer_edit_pose_meta)
                else:
                    tpl_retarget_pose_metas = get_retarget_pose(tpl_pose_meta0, refer_pose_meta, tpl_pose_metas, None, None)
            else:
               tpl_retarget_pose_metas = [AAPoseMeta.from_humanapi_meta(meta) for meta in tpl_pose_metas]

            cond_images = []
            for idx, meta in enumerate(tpl_retarget_pose_metas):
                if retarget_flag:
                    canvas = np.zeros_like(refer_img)
                    conditioning_image = draw_aapose_by_meta_new(canvas, meta)
                else:
                    canvas = np.zeros_like(frames[0])
                    conditioning_image = draw_aapose_by_meta_new(canvas, meta)
                    conditioning_image = padding_resize(conditioning_image, refer_img.shape[0], refer_img.shape[1])

                cond_images.append(conditioning_image)

            src_face_path = os.path.join(output_path, 'src_face.mp4')
            mpy.ImageSequenceClip(face_images, fps=fps).write_videofile(src_face_path, logger=None)

            src_pose_path = os.path.join(output_path, 'src_pose.mp4')
            mpy.ImageSequenceClip(cond_images, fps=fps).write_videofile(src_pose_path, logger=None)
            return True

    def get_editing_prompts(self, tpl_pose_metas, refer_pose_meta):
        arm_visible = False
        leg_visible = False
        for tpl_pose_meta in tpl_pose_metas:
            tpl_keypoints = tpl_pose_meta['keypoints_body']
            if tpl_keypoints[3].all() != 0 or tpl_keypoints[4].all() != 0 or tpl_keypoints[6].all() != 0 or tpl_keypoints[7].all() != 0:
                if (tpl_keypoints[3][0] <= 1 and tpl_keypoints[3][1] <= 1 and tpl_keypoints[3][2] >= 0.75) or (tpl_keypoints[4][0] <= 1 and tpl_keypoints[4][1] <= 1 and tpl_keypoints[4][2] >= 0.75) or \
                    (tpl_keypoints[6][0] <= 1 and tpl_keypoints[6][1] <= 1 and tpl_keypoints[6][2] >= 0.75) or (tpl_keypoints[7][0] <= 1 and tpl_keypoints[7][1] <= 1 and tpl_keypoints[7][2] >= 0.75):
                    arm_visible = True
            if tpl_keypoints[9].all() != 0 or tpl_keypoints[12].all() != 0 or tpl_keypoints[10].all() != 0 or tpl_keypoints[13].all() != 0:
                if (tpl_keypoints[9][0] <= 1 and tpl_keypoints[9][1] <= 1 and tpl_keypoints[9][2] >= 0.75) or (tpl_keypoints[12][0] <= 1 and tpl_keypoints[12][1] <= 1 and tpl_keypoints[12][2] >= 0.75) or \
                    (tpl_keypoints[10][0] <= 1 and tpl_keypoints[10][1] <= 1 and tpl_keypoints[10][2] >= 0.75) or (tpl_keypoints[13][0] <= 1 and tpl_keypoints[13][1] <= 1 and tpl_keypoints[13][2] >= 0.75):
                    leg_visible = True
            if arm_visible and leg_visible:
                break
        
        if leg_visible:
            if tpl_pose_meta['width'] > tpl_pose_meta['height']:
                tpl_prompt = "Change the person to a standard T-pose (facing forward with arms extended). The person is standing. Feet and Hands are visible in the image."
            else:
                tpl_prompt = "Change the person to a standard pose with the face oriented forward and arms extending straight down by the sides. The person is standing. Feet and Hands are visible in the image."

            if refer_pose_meta['width'] > refer_pose_meta['height']:
                refer_prompt = "Change the person to a standard T-pose (facing forward with arms extended). The person is standing. Feet and Hands are visible in the image."
            else:
                refer_prompt = "Change the person to a standard pose with the face oriented forward and arms extending straight down by the sides. The person is standing. Feet and Hands are visible in the image."
        elif arm_visible:
            if tpl_pose_meta['width'] > tpl_pose_meta['height']:
                tpl_prompt = "Change the person to a standard T-pose (facing forward with arms extended). Hands are visible in the image."
            else:
                tpl_prompt = "Change the person to a standard pose with the face oriented forward and arms extending straight down by the sides. Hands are visible in the image."

            if refer_pose_meta['width'] > refer_pose_meta['height']:
                refer_prompt = "Change the person to a standard T-pose (facing forward with arms extended). Hands are visible in the image."
            else:
                refer_prompt = "Change the person to a standard pose with the face oriented forward and arms extending straight down by the sides. Hands are visible in the image."
        else:
            tpl_prompt = "Change the person to face forward."
            refer_prompt = "Change the person to face forward."

        return tpl_prompt, refer_prompt
 
 
    def get_mask_from_face_bbox_v2(
        self,
        frames,
        pts_by_frame: dict[int, list[list[float]]] | None = None,
        lbs_by_frame: dict[int, list[int | float]] | None = None,
    ):
        """
        Args:
            frames: list/array of HxWx3 uint8 frames.
            pts_by_frame: {frame_idx: [[x,y], ...], ...}
            labels_by_frame: {frame_idx: [0/1,...], ...}

        Returns:
            all_mask: list[np.uint8 mask] of length len(frames), each (H, W) in {0,1}
        """

        print(f"lbs_by_frame:{lbs_by_frame}")
        print(f"pts_by_frame:{pts_by_frame}")

        # --- safety & normalization ---
        if pts_by_frame is None:
            pts_by_frame = {}
        if lbs_by_frame is None:
            lbs_by_frame = {}

        # normalize keys to int (in case they arrived as strings)
        pts_by_frame = {int(k): v for k, v in pts_by_frame.items()}
        lbs_by_frame = {int(k): v for k, v in lbs_by_frame.items()}

        H, W = frames[0].shape[:2]
        device = "cuda" if torch.cuda.is_available() else "cpu"

        with torch.autocast(device_type=device, dtype=torch.bfloat16 if device == "cuda" else torch.float16):
            # 1) init SAM2 video predictor state
            inference_state = self.predictor.init_state(images=np.array(frames), device=device)

            # 2) feed all per-frame clicks before propagating
            #    We use the *same obj_id* (0) so all clicks describe one object,
            #    no matter which frame they were added on.
            for fidx in sorted(pts_by_frame.keys()):
                pts = np.array(pts_by_frame.get(fidx, []), dtype=np.float32)
                lbs = np.array(lbs_by_frame.get(fidx, []), dtype=np.int32)

                if pts.size == 0:
                    continue  # nothing to add for this frame

                # (optional) sanity: make sure lens match
                if len(pts) != len(lbs):
                    raise ValueError(f"Points/labels length mismatch at frame {fidx}: {len(pts)} vs {len(lbs)}")

                self.predictor.add_new_points(
                    inference_state=inference_state,
                    frame_idx=int(fidx),
                    obj_id=0,
                    points=pts,
                    labels=lbs,
                )

            # 3) propagate across the whole video
            video_segments = {}
            for out_frame_idx, out_obj_ids, out_mask_logits in self.predictor.propagate_in_video(
                inference_state, start_frame_idx=0
            ):
                # store boolean masks per object id for this frame
                video_segments[out_frame_idx] = {
                    out_obj_id: (out_mask_logits[i] > 0.0).to("cpu").numpy()
                    for i, out_obj_id in enumerate(out_obj_ids)
                }

        # 4) collect masks in order; fall back to zeros where predictor returned nothing
        all_mask = []
        zero_mask = np.zeros((H, W), dtype=np.uint8)
        for out_frame_idx in range(len(frames)):
            if out_frame_idx in video_segments and len(video_segments[out_frame_idx]) > 0:
                mask = next(iter(video_segments[out_frame_idx].values()))
                if mask.ndim == 3:          # (1, H, W) -> (H, W)
                    mask = mask[0]
                mask = mask.astype(np.uint8)
            else:
                mask = zero_mask
            all_mask.append(mask)

        return all_mask



    def get_mask_from_face_bbox(self, frames, th_step, kp2ds_all):
        """
        Build masks using a face bounding box per key frame (derived from keypoints_face),
        then propagate with SAM2 across each chunk of frames.
        """
        H, W = frames[0].shape[:2]

        def _clip_box(x1, y1, x2, y2, W, H):
            x1 = max(0, min(int(x1), W - 1))
            x2 = max(0, min(int(x2), W - 1))
            y1 = max(0, min(int(y1), H - 1))
            y2 = max(0, min(int(y2), H - 1))
            if x2 <= x1: x2 = min(W - 1, x1 + 1)
            if y2 <= y1: y2 = min(H - 1, y1 + 1)
            return x1, y1, x2, y2

        frame_num = len(frames)
        if frame_num < th_step:
            num_step = 1
        else:
            num_step = (frame_num + th_step) // th_step

        all_mask = []

        for step_idx in range(num_step):
            each_frames = frames[step_idx * th_step:(step_idx + 1) * th_step]
            kp2ds = kp2ds_all[step_idx * th_step:(step_idx + 1) * th_step]
            if len(each_frames) == 0:
                continue

            # pick a few key frames in this chunk
            key_frame_num = 4 if len(each_frames) > 4 else 1
            key_frame_step = max(1, len(kp2ds) // key_frame_num)
            key_frame_index_list = list(range(0, len(kp2ds), key_frame_step))[:key_frame_num]

            # compute face boxes on the selected key frames
            key_frame_boxes = []
            for kfi in key_frame_index_list:
                meta = kp2ds[kfi]
                # get_face_bboxes returns (x1, x2, y1, y2) in your code
                x1, x2, y1, y2 = get_face_bboxes(
                    meta['keypoints_face'][:, :2],
                    scale=1.3,
                    image_shape=(H, W)
                )
                x1, y1, x2, y2 = _clip_box(x1, y1, x2, y2, W, H)
                key_frame_boxes.append(np.array([x1, y1, x2, y2], dtype=np.float32))

            # init SAM2 for this chunk
            with torch.autocast(device_type="cuda", dtype=torch.bfloat16):
                inference_state = self.predictor.init_state(images=np.array(each_frames), device="cuda")

                self.predictor.reset_state(inference_state)
                ann_obj_id = 1

                # seed with box prompts (preferred), else fall back to points
                for ann_frame_idx, box_xyxy in zip(key_frame_index_list, key_frame_boxes):
                    used_box = False
                    try:
                        # If your predictor exposes a box API, this is ideal.
                        _ = self.predictor.add_new_box(
                            inference_state=inference_state,
                            frame_idx=ann_frame_idx,
                            obj_id=ann_obj_id,
                            box=box_xyxy[None, :]  # shape (1, 4)
                        )
                        used_box = True
                    except Exception:
                        used_box = False

                    if not used_box:
                        # Fallback: sample a few positive points inside the box
                        x1, y1, x2, y2 = box_xyxy.astype(int)
                        cx, cy = (x1 + x2) // 2, (y1 + y2) // 2
                        pts = np.array([
                            [cx, cy],
                            [x1 + (x2 - x1) // 4, cy],
                            [x2 - (x2 - x1) // 4, cy],
                            [cx, y1 + (y2 - y1) // 4],
                            [cx, y2 - (y2 - y1) // 4],
                        ], dtype=np.int32)
                        labels = np.ones(len(pts), dtype=np.int32)  # 1 = positive
                        _ = self.predictor.add_new_points(
                            inference_state=inference_state,
                            frame_idx=ann_frame_idx,
                            obj_id=ann_obj_id,
                            points=pts,
                            labels=labels,
                        )

                # propagate across the chunk
                video_segments = {}
                for out_frame_idx, out_obj_ids, out_mask_logits in self.predictor.propagate_in_video(inference_state):
                    video_segments[out_frame_idx] = {
                        out_obj_id: (out_mask_logits[i] > 0.0).cpu().numpy()
                        for i, out_obj_id in enumerate(out_obj_ids)
                    }

            # collect masks (single object id)
            for out_frame_idx in range(len(video_segments)):
                # (H, W) boolean/uint8
                mask = next(iter(video_segments[out_frame_idx].values()))
                mask = mask[0].astype(np.uint8)
                all_mask.append(mask)

        return all_mask