[colmap] sparse reconstruction is converted to mvsnet format input

MVSNet File format

The project folder corresponding to each scenario should contain the following ：

.                          
├── images                 
│   ├── 00000000.jpg       
│   ├── 00000001.jpg       
│   └── ...                
├── cams                   
│   ├── 00000000_cam.txt   
│   ├── 00000001_cam.txt   
│   └── ...                
└── pair.txt               

You can use script files colmap2mvsnet.py take colmap The result of sparse reconstruction is mvsnet Input of format .

image file

All image files are stored in images In the folder , The index of the image is from 00000000 At the beginning 8 Digit number , The same index is used for the camera parameter file and the predicted output depth map .

Camera parameter file

The camera parameters of each image are stored in cam.txt In file , Camera parameter files include camera external parameters E=[R|t]、 Inside the camera K And depth range .

extrinsic
E00 E01 E02 E03
E10 E11 E12 E13
E20 E21 E22 E23
E30 E31 E32 E33

intrinsic
K00 K01 K02
K10 K11 K12
K20 K21 K22

DEPTH_MIN DEPTH_INTERVAL (DEPTH_NUM DEPTH_MAX)

Please note that the depth range and depth resolution are determined by the minimum depth DEPTH_MIN, Depth interval DEPTH_INTERVAL And the number of depth assumptions . in addition interval_scale Parameter is used to adjust the depth resolution , Therefore, the maximum depth value is ：

DEPTH_MAX = DEPTH_MIN + (interval_scale * DEPTH_INTERVAL) * (max_d - 1)

View selection file

The view selection results are stored in pair.txt In file , For each reference image , Calculate its view selection score with other views , And choose the one with the highest score 10 Views are stored in pair.txt In file ：

TOTAL_IMAGE_NUM
IMAGE_ID0                       # index of reference image 0 
10 ID0 SCORE0 ID1 SCORE1 ...    # 10 best source images for reference image 0 
IMAGE_ID1                       # index of reference image 1
10 ID0 SCORE0 ID1 SCORE1 ...    # 10 best source images for reference image 1 
...

Reference resources

• MVSNet Readme