Boese0601 / RC-MVSNet

[ECCV 2022] RC-MVSNet: Unsupervised Multi-View Stereo with Neural Rendering
https://boese0601.github.io/rc-mvsnet/
MIT License
210 stars 14 forks source link

Problem with reproducing "tanksandtemples" results #12

Open notmuchnerdy opened 1 year ago

notmuchnerdy commented 1 year ago

Hi,

I have reproduced the "tanksandtemples" .ply files according to the definition (without changing any parameter) on the main page of the RC-MVSNet. But unfortunately, I could not get meaningful results. I have uploaded a wide screenshot that depicts the results.

Results

You can see the depth estimates(left) and masks (right) on the left page. As you going to see here, Some of the depth estimates turn empty. Even worse, I could only get very few meaningful masks and not surprisingly I can only get the result like the right screen.

I would be appreciated it if you would help me to solve this issue.

Sincerely.

Boese0601 commented 1 year ago

What fusion did you use for point cloud generation? Did you use our provided fusion code and parameters, or other fusion e.g. gipuma fusion?

notmuchnerdy commented 1 year ago

I used your provided code "eval_rcmvsnet_tanks.py" and with its own threshold parameters as you see below:

scans = ['Family', 'Francis', 'Horse', 'Lighthouse', 'M60', 'Panther', 'Playground', 'Train']

    image_sizes = {'Family': (1920, 1080),
                        'Francis': (1920, 1080),
                        'Horse': (1920, 1080),
                        'Lighthouse': (2048, 1080),
                        'M60': (2048, 1080),
                        'Panther': (2048, 1080),
                        'Playground': (1920, 1080),
                        'Train': (1920, 1080)}
    geo_mask_thres = {'Family': 6,
                        'Francis': 8,
                        'Horse': 4,
                        'Lighthouse': 7,
                        'M60': 6,
                        'Panther': 7,
                        'Playground': 7,
                        'Train': 6} # num_consistency
    photo_thres = {'Family': 0.9,
                        'Francis': 0.8,
                        'Horse': 0.8,
                        'Lighthouse': 0.8,
                        'M60': 0.9,
                        'Panther': 0.9,
                        'Playground': 0.85,
                        'Train': 0.9} # prob_threshold
    geo_pixel_thres = {'Family': 0.75,
                        'Francis': 1.0,
                        'Horse': 1.25,
                        'Lighthouse': 1.0,
                        'M60': 0.75,
                        'Panther': 1.0,
                        'Playground': 1.0,
                        'Train': 1.5}    # img_dist_thresh

    geo_depth_thres = {'Family': 0.01,
                        'Francis': 0.01,
                        'Horse': 0.01,
                        'Lighthouse': 0.01,
                        'M60': 0.005,
                        'Panther': 0.01,
                        'Playground': 0.01,
                        'Train': 0.01}   # depth_thresh
hxmath commented 1 year ago

hi, @notmuchnerdy,I try to reproduce the results and run the eval_rcmvsnet_tanks.py, then I encounter the same problem with you.

lt-xiang commented 1 year ago

I used your provided code "eval_rcmvsnet_tanks.py" and with its own threshold parameters as you see below:

scans = ['Family', 'Francis', 'Horse', 'Lighthouse', 'M60', 'Panther', 'Playground', 'Train']

    image_sizes = {'Family': (1920, 1080),
                        'Francis': (1920, 1080),
                        'Horse': (1920, 1080),
                        'Lighthouse': (2048, 1080),
                        'M60': (2048, 1080),
                        'Panther': (2048, 1080),
                        'Playground': (1920, 1080),
                        'Train': (1920, 1080)}
    geo_mask_thres = {'Family': 6,
                        'Francis': 8,
                        'Horse': 4,
                        'Lighthouse': 7,
                        'M60': 6,
                        'Panther': 7,
                        'Playground': 7,
                        'Train': 6} # num_consistency
    photo_thres = {'Family': 0.9,
                        'Francis': 0.8,
                        'Horse': 0.8,
                        'Lighthouse': 0.8,
                        'M60': 0.9,
                        'Panther': 0.9,
                        'Playground': 0.85,
                        'Train': 0.9} # prob_threshold
    geo_pixel_thres = {'Family': 0.75,
                        'Francis': 1.0,
                        'Horse': 1.25,
                        'Lighthouse': 1.0,
                        'M60': 0.75,
                        'Panther': 1.0,
                        'Playground': 1.0,
                        'Train': 1.5}    # img_dist_thresh

    geo_depth_thres = {'Family': 0.01,
                        'Francis': 0.01,
                        'Horse': 0.01,
                        'Lighthouse': 0.01,
                        'M60': 0.005,
                        'Panther': 0.01,
                        'Playground': 0.01,
                        'Train': 0.01}   # depth_thresh

Hi, I met the same situation previouly. It is caused by depth range in the "cam.txt" of TAT.

1

RobinhoodKi commented 11 months ago

Same Problem

zhenni91 commented 4 months ago

i meet the same problem,how can i solve it