JeffreyChou commited on
Commit
74473e7
Β·
verified Β·
1 Parent(s): 965271a

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -7
README.md CHANGED
@@ -22,7 +22,7 @@ After decompression, please check the completeness of your download data and mak
22
  ```
23
  MM-AU # root of your MM-AU
24
  β”œβ”€β”€ CAP-DATA
25
- β”‚ β”œβ”€β”€ 1-10 #total 1556 video sequence
26
  β”‚ β”œβ”€β”€ 1
27
  β”‚ β”œβ”€β”€ 001537/images
28
  β”‚ β”œβ”€β”€ 000001.jpg
@@ -30,12 +30,12 @@ MM-AU # root of your MM-AU
30
  β”‚ β”œβ”€β”€ 2
31
  β”‚ β”œβ”€β”€ ......
32
  β”‚ β”œβ”€β”€ 10
33
- β”‚ β”œβ”€β”€ 11 #total 3083 video sequence
34
- β”‚ β”œβ”€β”€ 12-42 #total 1629 video sequence
35
- β”‚ β”œβ”€β”€ 43 #total 2150 video sequence
36
- β”‚ β”œβ”€β”€ 44-62 #total 1350 video sequence
37
  β”‚ β”œβ”€β”€ cap_text_annotations.xls
38
- β”œβ”€β”€ DADA-DATA #total 1962 video sequence
39
  β”‚ β”œβ”€β”€ 1
40
  β”‚ β”œβ”€β”€ 001/images
41
  β”‚ β”œβ”€β”€ 0001.png
@@ -50,7 +50,7 @@ NEW!: The coco style datasets for object detection task has been uploaded!
50
 
51
  Note: The object detection data used in the paper and the improved version MMAU-Detectv1 differ in both file names and number of videos due to different data cleaning methods and organization, but both maintain the same cocodataset style and the same dataset division strategy. The dataset used in the paper is provided to ensure the reproducibility of our paper, while the organization of MMAU-Detectv1 allows for better access to the video and image metadata when needed.
52
 
53
- Note: There are 295013 items in MMAU_det_paper/train, 64745 items in MMAU_det_paper/test, 62731 items in MMAU_det_paper/val, 299015 items in MMAU_detv1/train, 63473 items in MMAU_detv1/test and 65386 items in MMAU_detv1/val. After decompression, please check the completeness of your download data.
54
 
55
 
56
  NEW!:The checkpoints in the LOTVS-CAP Github repo has been uploaded!
 
22
  ```
23
  MM-AU # root of your MM-AU
24
  β”œβ”€β”€ CAP-DATA
25
+ β”‚ β”œβ”€β”€ 1-10 #total 1556 video sequences, takes 44GB
26
  β”‚ β”œβ”€β”€ 1
27
  β”‚ β”œβ”€β”€ 001537/images
28
  β”‚ β”œβ”€β”€ 000001.jpg
 
30
  β”‚ β”œβ”€β”€ 2
31
  β”‚ β”œβ”€β”€ ......
32
  β”‚ β”œβ”€β”€ 10
33
+ β”‚ β”œβ”€β”€ 11 #total 3083 video sequences, takes 96GB
34
+ β”‚ β”œβ”€β”€ 12-42 #total 1629 video sequences, takes 45GB
35
+ β”‚ β”œβ”€β”€ 43 #total 2150 video sequences, takes 44GB
36
+ β”‚ β”œβ”€β”€ 44-62 #total 1350 video sequences, takes 30GB
37
  β”‚ β”œβ”€β”€ cap_text_annotations.xls
38
+ β”œβ”€β”€ DADA-DATA #total 1962 video sequences, takes 131GB
39
  β”‚ β”œβ”€β”€ 1
40
  β”‚ β”œβ”€β”€ 001/images
41
  β”‚ β”œβ”€β”€ 0001.png
 
50
 
51
  Note: The object detection data used in the paper and the improved version MMAU-Detectv1 differ in both file names and number of videos due to different data cleaning methods and organization, but both maintain the same cocodataset style and the same dataset division strategy. The dataset used in the paper is provided to ensure the reproducibility of our paper, while the organization of MMAU-Detectv1 allows for better access to the video and image metadata when needed.
52
 
53
+ Note: There are 295013 items(48GB) in MMAU_det_paper/train, 64745 items(11GB) in MMAU_det_paper/test, 62731 items(9.9GB) in MMAU_det_paper/val, 299015(42GB) items in MMAU_detv1/train, 63473 items(9.0GB) in MMAU_detv1/test and 65386 items(9.1GB) in MMAU_detv1/val. After decompression, please check the completeness of your download data.
54
 
55
 
56
  NEW!:The checkpoints in the LOTVS-CAP Github repo has been uploaded!