File size: 27,002 Bytes
ffbe0b4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6f87787
ffbe0b4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0acbe25
ffbe0b4
 
 
 
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
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
from PIL.ImageOps import colorize, scale
import gradio as gr
import importlib
import sys
import os

from matplotlib.pyplot import step

from model_args import segtracker_args,sam_args,aot_args
from SegTracker import SegTracker

# sys.path.append('.')
# sys.path.append('..')

import cv2
from PIL import Image
from skimage.morphology.binary import binary_dilation
import argparse
import torch
import time
from seg_track_anything import aot_model2ckpt, tracking_objects_in_video, draw_mask
import gc
import numpy as np
import json
from tool.transfer_tools import mask2bbox

def clean():
    return None, None, None, None, None, None, [[], []]

def get_click_prompt(click_stack, point):

    click_stack[0].append(point["coord"])
    click_stack[1].append(point["mode"]
    )
    
    prompt = {
        "points_coord":click_stack[0],
        "points_mode":click_stack[1],
        "multimask":"True",
    }

    return prompt

def get_meta_from_video(input_video):
    if input_video is None:
        return None, None, None, ""

    print("get meta information of input video")
    cap = cv2.VideoCapture(input_video)
    
    _, first_frame = cap.read()
    cap.release()

    first_frame = cv2.cvtColor(first_frame, cv2.COLOR_BGR2RGB)

    return first_frame, first_frame, first_frame, ""

def get_meta_from_img_seq(input_img_seq):
    if input_img_seq is None:
        return None, None, None, ""

    print("get meta information of img seq")
    # Create dir
    file_name = input_img_seq.name.split('/')[-1].split('.')[0]
    file_path = f'./assets/{file_name}'
    if os.path.isdir(file_path):
        os.system(f'rm -r {file_path}')
    os.makedirs(file_path)
    # Unzip file
    os.system(f'unzip {input_img_seq.name} -d ./assets ')
    
    imgs_path = sorted([os.path.join(file_path, img_name) for img_name in os.listdir(file_path)])
    first_frame = imgs_path[0]
    first_frame = cv2.imread(first_frame)
    first_frame = cv2.cvtColor(first_frame, cv2.COLOR_BGR2RGB)

    return first_frame, first_frame, first_frame

def SegTracker_add_first_frame(Seg_Tracker, origin_frame, predicted_mask):
    with torch.cuda.amp.autocast():
        # Reset the first frame's mask
        frame_idx = 0
        Seg_Tracker.restart_tracker()
        Seg_Tracker.add_reference(origin_frame, predicted_mask, frame_idx)
        Seg_Tracker.first_frame_mask = predicted_mask

    return Seg_Tracker

def init_SegTracker(aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side, origin_frame):
    
    if origin_frame is None:
        return None, origin_frame, [[], []], ""

    # reset aot args
    aot_args["model"] = aot_model
    aot_args["model_path"] = aot_model2ckpt[aot_model]
    aot_args["long_term_mem_gap"] = long_term_mem
    aot_args["max_len_long_term"] = max_len_long_term
    # reset sam args
    segtracker_args["sam_gap"] = sam_gap
    segtracker_args["max_obj_num"] = max_obj_num
    sam_args["generator_args"]["points_per_side"] = points_per_side
    
    Seg_Tracker = SegTracker(segtracker_args, sam_args, aot_args)
    Seg_Tracker.restart_tracker()

    return Seg_Tracker, origin_frame, [[], []], ""

def init_SegTracker_Stroke(aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side, origin_frame):
    
    if origin_frame is None:
        return None, origin_frame, [[], []], origin_frame

    # reset aot args
    aot_args["model"] = aot_model
    aot_args["model_path"] = aot_model2ckpt[aot_model]
    aot_args["long_term_mem_gap"] = long_term_mem
    aot_args["max_len_long_term"] = max_len_long_term

    # reset sam args
    segtracker_args["sam_gap"] = sam_gap
    segtracker_args["max_obj_num"] = max_obj_num
    sam_args["generator_args"]["points_per_side"] = points_per_side
    
    Seg_Tracker = SegTracker(segtracker_args, sam_args, aot_args)
    Seg_Tracker.restart_tracker()
    return Seg_Tracker, origin_frame, [[], []], origin_frame

def undo_click_stack_and_refine_seg(Seg_Tracker, origin_frame, click_stack, aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side):
    
    if Seg_Tracker is None:
        return Seg_Tracker, origin_frame, [[], []]

    print("Undo!")
    if len(click_stack[0]) > 0:
        click_stack[0] = click_stack[0][: -1]
        click_stack[1] = click_stack[1][: -1]
    
    if len(click_stack[0]) > 0:
        prompt = {
            "points_coord":click_stack[0],
            "points_mode":click_stack[1],
            "multimask":"True",
        }

        masked_frame = seg_acc_click(Seg_Tracker, prompt, origin_frame)
        return Seg_Tracker, masked_frame, click_stack
    else:
        return Seg_Tracker, origin_frame, [[], []]


def seg_acc_click(Seg_Tracker, prompt, origin_frame):
    # seg acc to click
    predicted_mask, masked_frame = Seg_Tracker.seg_acc_click( 
                                                      origin_frame=origin_frame, 
                                                      coords=np.array(prompt["points_coord"]),
                                                      modes=np.array(prompt["points_mode"]),
                                                      multimask=prompt["multimask"],
                                                    )

    Seg_Tracker = SegTracker_add_first_frame(Seg_Tracker, origin_frame, predicted_mask)

    return masked_frame

def sam_click(Seg_Tracker, origin_frame, point_mode, click_stack, aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side, evt:gr.SelectData):
    """
    Args:
        origin_frame: nd.array
        click_stack: [[coordinate], [point_mode]]
    """

    print("Click")

    if point_mode == "Positive":
        point = {"coord": [evt.index[0], evt.index[1]], "mode": 1}
    else:
        # TODO:add everything positive points
        point = {"coord": [evt.index[0], evt.index[1]], "mode": 0}

    if Seg_Tracker is None:
        Seg_Tracker, _, _, _ = init_SegTracker(aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side, origin_frame)

    # get click prompts for sam to predict mask
    click_prompt = get_click_prompt(click_stack, point)

    # Refine acc to prompt
    masked_frame = seg_acc_click(Seg_Tracker, click_prompt, origin_frame)

    return Seg_Tracker, masked_frame, click_stack

def sam_stroke(Seg_Tracker, origin_frame, drawing_board, aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side):

    if Seg_Tracker is None:
        Seg_Tracker, _ , _, _ = init_SegTracker(aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side, origin_frame)
    
    print("Stroke")
    mask = drawing_board["mask"]
    bbox = mask2bbox(mask[:, :, 0])  # bbox: [[x0, y0], [x1, y1]]
    predicted_mask, masked_frame = Seg_Tracker.seg_acc_bbox(origin_frame, bbox)

    Seg_Tracker = SegTracker_add_first_frame(Seg_Tracker, origin_frame, predicted_mask)

    return Seg_Tracker, masked_frame, origin_frame

def gd_detect(Seg_Tracker, origin_frame, grounding_caption, box_threshold, text_threshold, aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side):
    if Seg_Tracker is None:
        Seg_Tracker, _ , _, _ = init_SegTracker(aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side, origin_frame)

    print("Detect")
    predicted_mask, annotated_frame= Seg_Tracker.detect_and_seg(origin_frame, grounding_caption, box_threshold, text_threshold)

    Seg_Tracker = SegTracker_add_first_frame(Seg_Tracker, origin_frame, predicted_mask)


    masked_frame = draw_mask(annotated_frame, predicted_mask)

    return Seg_Tracker, masked_frame, origin_frame

def segment_everything(Seg_Tracker, aot_model, long_term_mem, max_len_long_term, origin_frame, sam_gap, max_obj_num, points_per_side):
    
    if Seg_Tracker is None:
        Seg_Tracker, _ , _, _ = init_SegTracker(aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side, origin_frame)

    print("Everything")

    frame_idx = 0

    with torch.cuda.amp.autocast():
        pred_mask = Seg_Tracker.seg(origin_frame)
        torch.cuda.empty_cache()
        gc.collect()
        Seg_Tracker.add_reference(origin_frame, pred_mask, frame_idx)
        Seg_Tracker.first_frame_mask = pred_mask

    masked_frame = draw_mask(origin_frame.copy(), pred_mask)

    return Seg_Tracker, masked_frame

def add_new_object(Seg_Tracker):

    prev_mask = Seg_Tracker.first_frame_mask
    Seg_Tracker.update_origin_merged_mask(prev_mask)    
    Seg_Tracker.curr_idx += 1

    print("Ready to add new object!")

    return Seg_Tracker, [[], []]
## fps change to duration 4 argument
def tracking_objects(Seg_Tracker, input_video, input_img_seq, duration):
    print("Start tracking !")
    return tracking_objects_in_video(Seg_Tracker, input_video, input_img_seq, duration)

def seg_track_app():

    ##########################################################
    ######################  Front-end ########################
    ##########################################################
    app = gr.Blocks()

    with app:
        gr.Markdown(
            '''
            <div style="text-align:center;">
                <span style="font-size:3em; font-weight:bold;">Segment and Track for Swallowtail</span>
            </div>
            '''
        )

        click_stack = gr.State([[],[]]) # Storage clicks status
        origin_frame = gr.State(None)
        Seg_Tracker = gr.State(None)

        aot_model = gr.State(None)
        sam_gap = gr.State(None)
        points_per_side = gr.State(None)
        max_obj_num = gr.State(None)

        with gr.Row():
            # video input
            with gr.Column(scale=0.5):

                tab_video_input = gr.Tab(label="Video type input")
                with tab_video_input:
                    input_video = gr.Video(label='Input video').style(height=550)
                
                tab_img_seq_input = gr.Tab(label="Image-Seq type input")
                with tab_img_seq_input:
                    with gr.Row():
                        input_img_seq = gr.File(label='Input Image-Seq').style(height=550)
                        with gr.Column(scale=0.25):
                            extract_button = gr.Button(value="extract")
                            fps = gr.Slider(label='fps', minimum=5, maximum=50, value=8, step=1)

                input_first_frame = gr.Image(label='Segment result of first frame',interactive=True).style(height=550)


                tab_everything = gr.Tab(label="Everything")
                with tab_everything:
                    with gr.Row():
                        seg_every_first_frame = gr.Button(value="Segment everything for first frame", interactive=True)
                        point_mode = gr.Radio(
                            choices=["Positive"],
                            value="Positive",
                            label="Point Prompt",
                            interactive=True)

                        every_undo_but = gr.Button(
                                    value="Undo",
                                    interactive=True
                                    )

                            # every_reset_but = gr.Button(
                            #             value="Reset",
                            #             interactive=True
                            #                     )

                tab_click = gr.Tab(label="Click")
                with tab_click:
                    with gr.Row():
                        point_mode = gr.Radio(
                                    choices=["Positive",  "Negative"],
                                    value="Positive",
                                    label="Point Prompt",
                                    interactive=True)

                        # args for modify and tracking 
                        click_undo_but = gr.Button(
                                    value="Undo",
                                    interactive=True
                                    )
                            # click_reset_but = gr.Button(
                            #             value="Reset",
                            #             interactive=True
                            #                     )

                tab_stroke = gr.Tab(label="Stroke")
                with tab_stroke:
                    drawing_board = gr.Image(label='Drawing Board', tool="sketch", brush_radius=10, interactive=True)
                    with gr.Row():
                        seg_acc_stroke = gr.Button(value="Segment", interactive=True)
                        # stroke_reset_but = gr.Button(
                        #                 value="Reset",
                        #                 interactive=True
                        #                         )
                
                tab_text = gr.Tab(label="Text")
                with tab_text:
                    grounding_caption = gr.Textbox(label="Detection Prompt")
                    detect_button = gr.Button(value="Detect")
                    with gr.Accordion("Advanced options", open=False):
                        with gr.Row():
                            with gr.Column(scale=0.5):
                                box_threshold = gr.Slider(
                                    label="Box Threshold", minimum=0.0, maximum=1.0, value=0.25, step=0.001
                                )
                            with gr.Column(scale=0.5):
                                text_threshold = gr.Slider(
                                    label="Text Threshold", minimum=0.0, maximum=1.0, value=0.25, step=0.001
                                )

                with gr.Row():
                    with gr.Column(scale=0.5):
                        with gr.Tab(label="SegTracker Args"):
                            # args for tracking in video do segment-everthing
                            points_per_side = gr.Slider(
                                label = "points_per_side",
                                minimum= 1,
                                step = 1,
                                maximum=100,
                                value=16,
                                interactive=True
                            )

                            sam_gap = gr.Slider(
                                label='sam_gap',
                                minimum = 1,
                                step=1,
                                maximum = 9999,
                                value=100,
                                interactive=True,
                            )

                            max_obj_num = gr.Slider(
                                label='max_obj_num',
                                minimum = 50,
                                step=1,
                                maximum = 300,
                                value=255,
                                interactive=True
                            )
                            with gr.Accordion("aot advanced options", open=False):
                                aot_model = gr.Dropdown(
                                    label="aot_model",
                                    choices = [
                                        "deaotb",
                                        "deaotl",
                                        "r50_deaotl"
                                    ],
                                    value = "r50_deaotl",
                                    interactive=True,
                                )
                                long_term_mem = gr.Slider(label="long term memory gap", minimum=1, maximum=9999, value=9999, step=1)
                                max_len_long_term = gr.Slider(label="max len of long term memory", minimum=1, maximum=9999, value=9999, step=1)

                    with gr.Column():
                        new_object_button = gr.Button(
                            value="Add new object", 
                            interactive=True
                        )
                        reset_button = gr.Button(
                            value="Reset",
                            interactive=True,
                        )
                        track_for_video = gr.Button(
                            value="Start Tracking",
                                interactive=True,
                                )

            with gr.Column(scale=0.5):
                output_video = gr.Video(label='Output video').style(height=550)
                output_mask = gr.File(label="Predicted masks")

    ##########################################################
    ######################  back-end #########################
    ##########################################################

        # listen to the input_video to get the first frame of video
        input_video.change(
            fn=get_meta_from_video,
            inputs=[
                input_video
            ],
            outputs=[
                input_first_frame, origin_frame, drawing_board, grounding_caption
            ]
        )

        # listen to the input_img_seq to get the first frame of video
        input_img_seq.change(
            fn=get_meta_from_img_seq,
            inputs=[
                input_img_seq
            ],
            outputs=[
                input_first_frame, origin_frame, drawing_board, grounding_caption
            ]
        )
        
        #-------------- Input compont -------------
        tab_video_input.select(
            fn = clean,
            inputs=[],
            outputs=[
                input_video,
                input_img_seq,
                Seg_Tracker,
                input_first_frame,
                origin_frame,
                drawing_board,
                click_stack,
            ]
        )

        tab_img_seq_input.select(
            fn = clean,
            inputs=[],
            outputs=[
                input_video,
                input_img_seq,
                Seg_Tracker,
                input_first_frame,
                origin_frame,
                drawing_board,
                click_stack,
            ]
        )

        extract_button.click(
            fn=get_meta_from_img_seq,
            inputs=[
                input_img_seq
            ],
            outputs=[
                input_first_frame, origin_frame, drawing_board
            ]
        )


        # ------------------- Interactive component -----------------

        # listen to the tab to init SegTracker
        tab_everything.select(
            fn=init_SegTracker,
            inputs=[
                aot_model,
                long_term_mem,
                max_len_long_term, 
                sam_gap,
                max_obj_num,
                points_per_side,
                origin_frame
            ],
            outputs=[
                Seg_Tracker, input_first_frame, click_stack, grounding_caption
            ],
            queue=False,
            
        )
        
        tab_click.select(
            fn=init_SegTracker,
            inputs=[
                aot_model,
                long_term_mem,
                max_len_long_term,
                sam_gap,
                max_obj_num,
                points_per_side,
                origin_frame
            ],
            outputs=[
                Seg_Tracker, input_first_frame, click_stack, grounding_caption
            ],
            queue=False,
        )

        tab_stroke.select(
            fn=init_SegTracker_Stroke,
            inputs=[
                aot_model,
                long_term_mem,
                max_len_long_term,
                sam_gap,
                max_obj_num,
                points_per_side,
                origin_frame,
            ],
            outputs=[
                Seg_Tracker, input_first_frame, click_stack, drawing_board
            ],
            queue=False,
        )

        tab_text.select(
            fn=init_SegTracker,
            inputs=[
                aot_model,
                long_term_mem,
                max_len_long_term,
                sam_gap,
                max_obj_num,
                points_per_side,
                origin_frame
            ],
            outputs=[
                Seg_Tracker, input_first_frame, click_stack, grounding_caption
            ],
            queue=False,
        )

        # Use SAM to segment everything for the first frame of video
        seg_every_first_frame.click(
            fn=segment_everything,
            inputs=[
                Seg_Tracker,
                aot_model,
                long_term_mem,
                max_len_long_term,
                origin_frame,
                sam_gap,
                max_obj_num,
                points_per_side,

            ],
            outputs=[
                Seg_Tracker,
                input_first_frame,
            ],
            )
        
        # Interactively modify the mask acc click
        input_first_frame.select(
            fn=sam_click,
            inputs=[
                Seg_Tracker, origin_frame, point_mode, click_stack,
                aot_model,
                long_term_mem,
                max_len_long_term,
                sam_gap,
                max_obj_num,
                points_per_side,
            ],
            outputs=[
                Seg_Tracker, input_first_frame, click_stack
            ]
        )

        # Interactively segment acc stroke
        seg_acc_stroke.click(
            fn=sam_stroke,
            inputs=[
                Seg_Tracker, origin_frame, drawing_board,
                aot_model,
                long_term_mem,
                max_len_long_term,
                sam_gap,
                max_obj_num,
                points_per_side,
            ],
            outputs=[
                Seg_Tracker, input_first_frame, drawing_board
            ]
        )

        # Use grounding-dino to detect object
        detect_button.click(
            fn=gd_detect, 
            inputs=[
                Seg_Tracker, origin_frame, grounding_caption, box_threshold, text_threshold,
                aot_model, long_term_mem, max_len_long_term, sam_gap, max_obj_num, points_per_side
                ], 
            outputs=[
                Seg_Tracker, input_first_frame
                ]
                )

        # Add new object
        new_object_button.click(
            fn=add_new_object,
            inputs=
            [
                Seg_Tracker
            ],
            outputs=
            [
                Seg_Tracker, click_stack
            ]
        )

        # Track object in video
        track_for_video.click(
            fn=tracking_objects,
            inputs=[
                Seg_Tracker,
                input_video,
                input_img_seq,
                fps,
            ],
            outputs=[
                output_video, output_mask
            ]
        )

        # ----------------- Reset and Undo ---------------------------

        # Rest 
        reset_button.click(
            fn=init_SegTracker,
            inputs=[
                aot_model,
                long_term_mem,
                max_len_long_term,
                sam_gap,
                max_obj_num,
                points_per_side,
                origin_frame
            ],
            outputs=[
                Seg_Tracker, input_first_frame, click_stack, grounding_caption
            ],
            queue=False,
            show_progress=False
        ) 

        # every_reset_but.click(
        #     fn=init_SegTracker,
        #     inputs=[
        #         aot_model,
        #         sam_gap,
        #         max_obj_num,
        #         points_per_side,
        #         origin_frame
        #     ],
        #     outputs=[
        #         Seg_Tracker, input_first_frame, click_stack, grounding_caption
        #     ],
        #     queue=False,
        #     show_progress=False
        # ) 

        # click_reset_but.click(
        #     fn=init_SegTracker,
        #     inputs=[
        #         aot_model,
        #         sam_gap,
        #         max_obj_num,
        #         points_per_side,
        #         origin_frame
        #     ],
        #     outputs=[
        #         Seg_Tracker, input_first_frame, click_stack, grounding_caption
        #     ],
        #     queue=False,
        #     show_progress=False
        # ) 

        # stroke_reset_but.click(
        #     fn=init_SegTracker_Stroke,
        #     inputs=[
        #         aot_model,
        #         sam_gap,
        #         max_obj_num,
        #         points_per_side,
        #         origin_frame,
        #     ],
        #     outputs=[
        #         Seg_Tracker, input_first_frame, click_stack, drawing_board
        #     ],
        #     queue=False,
        #     show_progress=False
        # )

        # Undo click
        click_undo_but.click(
            fn = undo_click_stack_and_refine_seg,
            inputs=[
                Seg_Tracker, origin_frame, click_stack,
                aot_model,
                long_term_mem,
                max_len_long_term,
                sam_gap,
                max_obj_num,
                points_per_side,
            ],
            outputs=[
               Seg_Tracker, input_first_frame, click_stack
            ]
        )

        every_undo_but.click(
            fn = undo_click_stack_and_refine_seg,
            inputs=[
                Seg_Tracker, origin_frame, click_stack,
                aot_model,
                long_term_mem,
                max_len_long_term,
                sam_gap,
                max_obj_num,
                points_per_side,
            ],
            outputs=[
               Seg_Tracker, input_first_frame, click_stack
            ]
        )
        
        with gr.Tab(label='Video example'):
            gr.Examples(
                examples=[
                    # os.path.join(os.path.dirname(__file__), "assets", "840_iSXIa0hE8Ek.mp4"),
                    os.path.join(os.path.dirname(__file__), "assets", "blackswan.mp4"),
                    #os.path.join(os.path.dirname(__file__), "assets", "bear.mp4"),
                    # os.path.join(os.path.dirname(__file__), "assets", "camel.mp4"),
                    # os.path.join(os.path.dirname(__file__), "assets", "skate-park.mp4"),
                    # os.path.join(os.path.dirname(__file__), "assets", "swing.mp4"),
                    ],
                inputs=[input_video],
            )
        
        with gr.Tab(label='Image-seq expamle'):
            gr.Examples(
                examples=[
                    os.path.join(os.path.dirname(__file__), "assets", "840_iSXIa0hE8Ek.zip"),
                ],
                inputs=[input_img_seq],
            )
    
    app.queue(concurrency_count=1)
    app.launch(debug=True, enable_queue=True)


if __name__ == "__main__":
    seg_track_app()