找不到数据集

[sunqimin]

您好，我执行test出现了下面这个问题，但是我查看了OSTrack文件夹里面并没有包含data这个文件，所以想请教您我接下来该怎么做 OSError: D:/SQM/tracking/OSTrack/data/lasot/airplane/airplane-1/groundtruth.txt　not found.

[franktpmvu]

download dataset and put it on correct path

see:

https://github.com/botaoye/OSTrack

[franktpmvu]

因為你的測試還沒執行完畢 當每支影片執行完畢時會列印該影片的FPS 當所有影片都執行完畢時會列印所有影片的平均FPS sunqimin @.***> 於 2024年4月9日 週二 上午9:37寫道：

按照您所说的我下好了数据，由于我的服务器带不动，所以我把threads改成了6，num_gpus改成了2，然后执行test得到了如下的结果，请问其中没有FPS的信息是因为什么呢 $ sh test.sh Evaluating 1 trackers on 280 sequences Tracker: ostrack vitb_384_mae_ce_32x4_ep300_neighbor None , Sequence: book-3 test config: {'MODEL': {'PRETRAIN_FILE': 'mae_pretrain_vit_base.pth', 'EXTRA_ME RGER': False, 'RETURN_INTER': False, 'RETURN_STAGES': [2, 5, 8, 11], 'BACKBONE': {'TYPE': 'vit_base_patch16_224_ce', 'STRIDE': 16, 'MID_PE': False, 'SEP_SEG': F alse, 'CAT_MODE': 'direct', 'MERGE_LAYER': 0, 'ADD_CLS_TOKEN': False, 'CLS_TOKEN _USE_MODE': 'ignore', 'CE_LOC': [3, 6, 9], 'CE_KEEP_RATIO': [0.7, 0.7, 0.7], 'CE _TEMPLATE_RANGE': 'CTR_POINT'}, 'HEAD': {'TYPE': 'CENTER_NEIGHBOR', 'NUM_CHANNEL S': 256}}, 'TRAIN': {'LR': 0.0004, 'WEIGHT_DECAY': 0.0001, 'EPOCH': 300, 'LR_DRO P_EPOCH': 240, 'BATCH_SIZE': 32, 'NUM_WORKER': 10, 'OPTIMIZER': 'ADAMW', 'BACKBO NE_MULTIPLIER': 0.1, 'GIOU_WEIGHT': 2.0, 'L1_WEIGHT': 5.0, 'FREEZE_LAYERS': [0], 'PRINT_INTERVAL': 50, 'VAL_EPOCH_INTERVAL': 20, 'GRAD_CLIP_NORM': 0.1, 'AMP': F alse, 'CE_START_EPOCH': 20, 'CE_WARM_EPOCH': 80, 'DROP_PATH_RATE': 0.1, 'SCHEDUL ER': {'TYPE': 'step', 'DECAY_RATE': 0.1}}, 'DATA': {'SAMPLER_MODE': 'causal', 'M EAN': [0.485, 0.456, 0.406], 'STD': [0.229, 0.224, 0.225], 'MAX_SAMPLE_INTERVAL' : 200, 'TRAIN': {'DATASETS_NAME': ['LASOT', 'GOT10K_vottrain', 'COCO17', 'TRACKI NGNET'], 'DATASETS_RATIO': [1, 1, 1, 1], 'SAMPLE_PER_EPOCH': 60000}, 'VAL': {'DA TASETS_NAME': ['GOT10K_votval'], 'DATASETS_RATIO': [1], 'SAMPLE_PER_EPOCH': 1000 0}, 'SEARCH': {'SIZE': 384, 'FACTOR': 5.0, 'CENTER_JITTER': 4.5, 'SCALE_JITTER': 0.5, 'NUMBER': 1}, 'TEMPLATE': {'NUMBER': 1, 'SIZE': 192, 'FACTOR': 2.0, 'CENTE R_JITTER': 0, 'SCALE_JITTER': 0}}, 'TEST': {'TEMPLATE_FACTOR': 2.0, 'TEMPLATE_SI ZE': 192, 'SEARCH_FACTOR': 5.0, 'SEARCH_SIZE': 384, 'EPOCH': 300}} Tracker: ostrack vitb_384_mae_ce_32x4_ep300_neighbor None , Sequence: car-2 test config: {'MODEL': {'PRETRAIN_FILE': 'mae_pretrain_vit_base.pth', 'EXTRA_ME RGER': False, 'RETURN_INTER': False, 'RETURN_STAGES': [2, 5, 8, 11], 'BACKBONE': {'TYPE': 'vit_base_patch16_224_ce', 'STRIDE': 16, 'MID_PE': False, 'SEP_SEG': F alse, 'CAT_MODE': 'direct', 'MERGE_LAYER': 0, 'ADD_CLS_TOKEN': False, 'CLS_TOKEN _USE_MODE': 'ignore', 'CE_LOC': [3, 6, 9], 'CE_KEEP_RATIO': [0.7, 0.7, 0.7], 'CE _TEMPLATE_RANGE': 'CTR_POINT'}, 'HEAD': {'TYPE': 'CENTER_NEIGHBOR', 'NUM_CHANNEL S': 256}}, 'TRAIN': {'LR': 0.0004, 'WEIGHT_DECAY': 0.0001, 'EPOCH': 300, 'LR_DRO P_EPOCH': 240, 'BATCH_SIZE': 32, 'NUM_WORKER': 10, 'OPTIMIZER': 'ADAMW', 'BACKBO NE_MULTIPLIER': 0.1, 'GIOU_WEIGHT': 2.0, 'L1_WEIGHT': 5.0, 'FREEZE_LAYERS': [0], 'PRINT_INTERVAL': 50, 'VAL_EPOCH_INTERVAL': 20, 'GRAD_CLIP_NORM': 0.1, 'AMP': F alse, 'CE_START_EPOCH': 20, 'CE_WARM_EPOCH': 80, 'DROP_PATH_RATE': 0.1, 'SCHEDUL ER': {'TYPE': 'step', 'DECAY_RATE': 0.1}}, 'DATA': {'SAMPLER_MODE': 'causal', 'M EAN': [0.485, 0.456, 0.406], 'STD': [0.229, 0.224, 0.225], 'MAX_SAMPLE_INTERVAL' : 200, 'TRAIN': {'DATASETS_NAME': ['LASOT', 'GOT10K_vottrain', 'COCO17', 'TRACKI NGNET'], 'DATASETS_RATIO': [1, 1, 1, 1], 'SAMPLE_PER_EPOCH': 60000}, 'VAL': {'DA TASETS_NAME': ['GOT10K_votval'], 'DATASETS_RATIO': [1], 'SAMPLE_PER_EPOCH': 1000 0}, 'SEARCH': {'SIZE': 384, 'FACTOR': 5.0, 'CENTER_JITTER': 4.5, 'SCALE_JITTER': 0.5, 'NUMBER': 1}, 'TEMPLATE': {'NUMBER': 1, 'SIZE': 192, 'FACTOR': 2.0, 'CENTE R_JITTER': 0, 'SCALE_JITTER': 0}}, 'TEST': {'TEMPLATE_FACTOR': 2.0, 'TEMPLATE_SI ZE': 192, 'SEARCH_FACTOR': 5.0, 'SEARCH_SIZE': 384, 'EPOCH': 300}} Tracker: ostrack vitb_384_mae_ce_32x4_ep300_neighbor None , Sequence: surfboard -12 Tracker: ostrack vitb_384_mae_ce_32x4_ep300_neighbor None , Sequence: airplane- 1 test config: {'MODEL': {'PRETRAIN_FILE': 'mae_pretrain_vit_base.pth', 'EXTRA_ME RGER': False, 'RETURN_INTER': False, 'RETURN_STAGES': [2, 5, 8, 11], 'BACKBONE': {'TYPE': 'vit_base_patch16_224_ce', 'STRIDE': 16, 'MID_PE': False, 'SEP_SEG': F alse, 'CAT_MODE': 'direct', 'MERGE_LAYER': 0, 'ADD_CLS_TOKEN': False, 'CLS_TOKEN _USE_MODE': 'ignore', 'CE_LOC': [3, 6, 9], 'CE_KEEP_RATIO': [0.7, 0.7, 0.7], 'CE _TEMPLATE_RANGE': 'CTR_POINT'}, 'HEAD': {'TYPE': 'CENTER_NEIGHBOR', 'NUM_CHANNEL S': 256}}, 'TRAIN': {'LR': 0.0004, 'WEIGHT_DECAY': 0.0001, 'EPOCH': 300, 'LR_DRO P_EPOCH': 240, 'BATCH_SIZE': 32, 'NUM_WORKER': 10, 'OPTIMIZER': 'ADAMW', 'BACKBO NE_MULTIPLIER': 0.1, 'GIOU_WEIGHT': 2.0, 'L1_WEIGHT': 5.0, 'FREEZE_LAYERS': [0], 'PRINT_INTERVAL': 50, 'VAL_EPOCH_INTERVAL': 20, 'GRAD_CLIP_NORM': 0.1, 'AMP': F alse, 'CE_START_EPOCH': 20, 'CE_WARM_EPOCH': 80, 'DROP_PATH_RATE': 0.1, 'SCHEDUL ER': {'TYPE': 'step', 'DECAY_RATE': 0.1}}, 'DATA': {'SAMPLER_MODE': 'causal', 'M EAN': [0.485, 0.456, 0.406], 'STD': [0.229, 0.224, 0.225], 'MAX_SAMPLE_INTERVAL' : 200, 'TRAIN': {'DATASETS_NAME': ['LASOT', 'GOT10K_vottrain', 'COCO17', 'TRACKI NGNET'], 'DATASETS_RATIO': [1, 1, 1, 1], 'SAMPLE_PER_EPOCH': 60000}, 'VAL': {'DA TASETS_NAME': ['GOT10K_votval'], 'DATASETS_RATIO': [1], 'SAMPLE_PER_EPOCH': 1000 0}, 'SEARCH': {'SIZE': 384, 'FACTOR': 5.0, 'CENTER_JITTER': 4.5, 'SCALE_JITTER': 0.5, 'NUMBER': 1}, 'TEMPLATE': {'NUMBER': 1, 'SIZE': 192, 'FACTOR': 2.0, 'CENTE R_JITTER': 0, 'SCALE_JITTER': 0}}, 'TEST': {'TEMPLATE_FACTOR': 2.0, 'TEMPLATE_SI ZE': 192, 'SEARCH_FACTOR': 5.0, 'SEARCH_SIZE': 384, 'EPOCH': 300}} Tracker: ostrack vitb_384_mae_ce_32x4_ep300_neighbor None , Sequence: spider-14 Tracker: ostrack vitb_384_mae_ce_32x4_ep300_neighbor None , Sequence: bicycle-2 test config: {'MODEL': {'PRETRAIN_FILE': 'mae_pretrain_vit_base.pth', 'EXTRA_ME RGER': False, 'RETURN_INTER': False, 'RETURN_STAGES': [2, 5, 8, 11], 'BACKBONE': {'TYPE': 'vit_base_patch16_224_ce', 'STRIDE': 16, 'MID_PE': False, 'SEP_SEG': F alse, 'CAT_MODE': 'direct', 'MERGE_LAYER': 0, 'ADD_CLS_TOKEN': False, 'CLS_TOKEN _USE_MODE': 'ignore', 'CE_LOC': [3, 6, 9], 'CE_KEEP_RATIO': [0.7, 0.7, 0.7], 'CE _TEMPLATE_RANGE': 'CTR_POINT'}, 'HEAD': {'TYPE': 'CENTER_NEIGHBOR', 'NUM_CHANNEL S': 256}}, 'TRAIN': {'LR': 0.0004, 'WEIGHT_DECAY': 0.0001, 'EPOCH': 300, 'LR_DRO P_EPOCH': 240, 'BATCH_SIZE': 32, 'NUM_WORKER': 10, 'OPTIMIZER': 'ADAMW', 'BACKBO NE_MULTIPLIER': 0.1, 'GIOU_WEIGHT': 2.0, 'L1_WEIGHT': 5.0, 'FREEZE_LAYERS': [0], 'PRINT_INTERVAL': 50, 'VAL_EPOCH_INTERVAL': 20, 'GRAD_CLIP_NORM': 0.1, 'AMP': F alse, 'CE_START_EPOCH': 20, 'CE_WARM_EPOCH': 80, 'DROP_PATH_RATE': 0.1, 'SCHEDUL ER': {'TYPE': 'step', 'DECAY_RATE': 0.1}}, 'DATA': {'SAMPLER_MODE': 'causal', 'M EAN': [0.485, 0.456, 0.406], 'STD': [0.229, 0.224, 0.225], 'MAX_SAMPLE_INTERVAL' : 200, 'TRAIN': {'DATASETS_NAME': ['LASOT', 'GOT10K_vottrain', 'COCO17', 'TRACKI NGNET'], 'DATASETS_RATIO': [1, 1, 1, 1], 'SAMPLE_PER_EPOCH': 60000}, 'VAL': {'DA TASETS_NAME': ['GOT10K_votval'], 'DATASETS_RATIO': [1], 'SAMPLE_PER_EPOCH': 1000 0}, 'SEARCH': {'SIZE': 384, 'FACTOR': 5.0, 'CENTER_JITTER': 4.5, 'SCALE_JITTER': 0.5, 'NUMBER': 1}, 'TEMPLATE': {'NUMBER': 1, 'SIZE': 192, 'FACTOR': 2.0, 'CENTE R_JITTER': 0, 'SCALE_JITTER': 0}}, 'TEST': {'TEMPLATE_FACTOR': 2.0, 'TEMPLATE_SI ZE': 192, 'SEARCH_FACTOR': 5.0, 'SEARCH_SIZE': 384, 'EPOCH': 300}}

— Reply to this email directly, view it on GitHub https://github.com/franktpmvu/NeighborTrack/issues/8#issuecomment-2043988566, or unsubscribe https://github.com/notifications/unsubscribe-auth/AJJLQVYMY5VMDUOINZLLIXTY4NA5DAVCNFSM6AAAAABFS3UYQ6VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANBTHE4DQNJWGY . You are receiving this because you commented.Message ID: @.***>

[sunqimin]

感谢您的回复！我后面得到了test的结果，请问跟踪结果可以可视化吗？如果可以的话，那么如何可视化呢？

[franktpmvu]

你可以用video_test.sh來可視化你的video

[sunqimin]

After I applied neighbortrack to other trackers, the results on lasot got worse, and the performance on uav got better. I haven't found the reason, perhaps you know why

[franktpmvu]

I think you can analyze both dataset's occluder, in my hypothesis, lasot dataset's occluder object is the other same class object, and UAV dataset is the background, neighbortrack also uses an appearance-based tracker, so if the occluder has a similar appearance, this case is more difficult than occluded by un-similar background. but it is based on your base tracker, not an attribute of neighbortrack, in my hypothesis, if you are not using an appearance-based tracker, then this phenomenon will not Appear.

[sunqimin]

Thank you very much! The base trackers I use are odtrack and hiptrack. I'm not sure how to solve this problem, do you have any good suggestions?