File size: 7,888 Bytes
04daa95
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from statistics import mode
import torch
import torch.nn.functional as F
import os
import sys
sys.path.append("./aot")
from aot.networks.engines.aot_engine import AOTEngine,AOTInferEngine
from aot.networks.engines.deaot_engine import DeAOTEngine,DeAOTInferEngine
import importlib
import numpy as np
from PIL import Image
from skimage.morphology.binary import binary_dilation


np.random.seed(200)
_palette = ((np.random.random((3*255))*0.7+0.3)*255).astype(np.uint8).tolist()
_palette = [0,0,0]+_palette

import aot.dataloaders.video_transforms as tr
from aot.utils.checkpoint import load_network
from aot.networks.models import build_vos_model
from aot.networks.engines import build_engine
from torchvision import transforms

class AOTTracker(object):
    def __init__(self, cfg, gpu_id=0):
        self.gpu_id = gpu_id
        self.model = build_vos_model(cfg.MODEL_VOS, cfg).cuda(gpu_id)
        self.model, _ = load_network(self.model, cfg.TEST_CKPT_PATH, gpu_id)
        # self.engine = self.build_tracker_engine(cfg.MODEL_ENGINE,
        #                            aot_model=self.model,
        #                            gpu_id=gpu_id,
        #                            short_term_mem_skip=4,
        #                            long_term_mem_gap=cfg.TEST_LONG_TERM_MEM_GAP)
        self.engine = build_engine(cfg.MODEL_ENGINE,
                                   phase='eval',
                                   aot_model=self.model,
                                   gpu_id=gpu_id,
                                   short_term_mem_skip=1,
                                   long_term_mem_gap=cfg.TEST_LONG_TERM_MEM_GAP,
                                   max_len_long_term=cfg.MAX_LEN_LONG_TERM)
       
        self.transform = transforms.Compose([
            tr.MultiRestrictSize(cfg.TEST_MAX_SHORT_EDGE,
                                 cfg.TEST_MAX_LONG_EDGE, cfg.TEST_FLIP, 
                                 cfg.TEST_MULTISCALE, cfg.MODEL_ALIGN_CORNERS),
            tr.MultiToTensor()
        ])

        self.model.eval()

    @torch.no_grad()
    def add_reference_frame(self, frame, mask, obj_nums, frame_step, incremental=False):
        # mask = cv2.resize(mask, frame.shape[:2][::-1], interpolation = cv2.INTER_NEAREST)

        sample = {
            'current_img': frame,
            'current_label': mask,
        }
    
        sample = self.transform(sample)
        frame = sample[0]['current_img'].unsqueeze(0).float().cuda(self.gpu_id)
        mask = sample[0]['current_label'].unsqueeze(0).float().cuda(self.gpu_id)
        _mask = F.interpolate(mask,size=frame.shape[-2:],mode='nearest')

        if incremental:
            self.engine.add_reference_frame_incremental(frame, _mask, obj_nums=obj_nums, frame_step=frame_step)
        else:
            self.engine.add_reference_frame(frame, _mask, obj_nums=obj_nums, frame_step=frame_step)



    @torch.no_grad()
    def track(self, image):
        output_height, output_width = image.shape[0], image.shape[1]
        sample = {'current_img': image}
        sample = self.transform(sample)
        image = sample[0]['current_img'].unsqueeze(0).float().cuda(self.gpu_id)
        self.engine.match_propogate_one_frame(image)
        pred_logit = self.engine.decode_current_logits((output_height, output_width))

        # pred_prob = torch.softmax(pred_logit, dim=1)
        pred_label = torch.argmax(pred_logit, dim=1,
                                    keepdim=True).float()

        return  pred_label
    
    @torch.no_grad()
    def update_memory(self, pred_label):
        self.engine.update_memory(pred_label)
    
    @torch.no_grad()
    def restart(self):
        self.engine.restart_engine()
    
    @torch.no_grad()
    def build_tracker_engine(self, name, **kwargs):
        if name == 'aotengine':
            return AOTTrackerInferEngine(**kwargs)
        elif name == 'deaotengine':
            return DeAOTTrackerInferEngine(**kwargs)
        else:
            raise NotImplementedError


class AOTTrackerInferEngine(AOTInferEngine):
    def __init__(self, aot_model, gpu_id=0, long_term_mem_gap=9999, short_term_mem_skip=1, max_aot_obj_num=None):
        super().__init__(aot_model, gpu_id, long_term_mem_gap, short_term_mem_skip, max_aot_obj_num)
    def add_reference_frame_incremental(self, img, mask, obj_nums, frame_step=-1):
        if isinstance(obj_nums, list):
            obj_nums = obj_nums[0]
        self.obj_nums = obj_nums
        aot_num = max(np.ceil(obj_nums / self.max_aot_obj_num), 1)
        while (aot_num > len(self.aot_engines)):
            new_engine = AOTEngine(self.AOT, self.gpu_id,
                                   self.long_term_mem_gap,
                                   self.short_term_mem_skip)
            new_engine.eval()
            self.aot_engines.append(new_engine)

        separated_masks, separated_obj_nums = self.separate_mask(
            mask, obj_nums)
        img_embs = None
        for aot_engine, separated_mask, separated_obj_num in zip(
                self.aot_engines, separated_masks, separated_obj_nums):
            if aot_engine.obj_nums is None or aot_engine.obj_nums[0] < separated_obj_num:
                aot_engine.add_reference_frame(img,
                                            separated_mask,
                                            obj_nums=[separated_obj_num],
                                            frame_step=frame_step,
                                            img_embs=img_embs)
            else:
                aot_engine.update_short_term_memory(separated_mask)
                
            if img_embs is None:  # reuse image embeddings
                img_embs = aot_engine.curr_enc_embs

        self.update_size()



class DeAOTTrackerInferEngine(DeAOTInferEngine):
    def __init__(self, aot_model, gpu_id=0, long_term_mem_gap=9999, short_term_mem_skip=1, max_aot_obj_num=None):
        super().__init__(aot_model, gpu_id, long_term_mem_gap, short_term_mem_skip, max_aot_obj_num)
    def add_reference_frame_incremental(self, img, mask, obj_nums, frame_step=-1):
        if isinstance(obj_nums, list):
            obj_nums = obj_nums[0]
        self.obj_nums = obj_nums
        aot_num = max(np.ceil(obj_nums / self.max_aot_obj_num), 1)
        while (aot_num > len(self.aot_engines)):
            new_engine = DeAOTEngine(self.AOT, self.gpu_id,
                                   self.long_term_mem_gap,
                                   self.short_term_mem_skip)
            new_engine.eval()
            self.aot_engines.append(new_engine)

        separated_masks, separated_obj_nums = self.separate_mask(
            mask, obj_nums)
        img_embs = None
        for aot_engine, separated_mask, separated_obj_num in zip(
                self.aot_engines, separated_masks, separated_obj_nums):
            if aot_engine.obj_nums is None or aot_engine.obj_nums[0] < separated_obj_num:
                aot_engine.add_reference_frame(img,
                                            separated_mask,
                                            obj_nums=[separated_obj_num],
                                            frame_step=frame_step,
                                            img_embs=img_embs)
            else:
                aot_engine.update_short_term_memory(separated_mask)
                
            if img_embs is None:  # reuse image embeddings
                img_embs = aot_engine.curr_enc_embs

        self.update_size()


def get_aot(args):
    # build vos engine
    engine_config = importlib.import_module('configs.' + 'pre_ytb_dav')
    cfg = engine_config.EngineConfig(args['phase'], args['model'])
    cfg.TEST_CKPT_PATH = args['model_path']
    cfg.TEST_LONG_TERM_MEM_GAP = args['long_term_mem_gap']
    cfg.MAX_LEN_LONG_TERM = args['max_len_long_term']
    # init AOTTracker
    tracker = AOTTracker(cfg, args['gpu_id'])
    return tracker