Monocular Depth Estimation Challenge

Monocular Depth Estimation Challenge Workshop at WACV 2023

[Monocular Depth Estimation Challenge @DepthChallenge]

The #MDEC (Monocular Depth Estimation Challenge) for our #WACV23 workshop will be going live on October 5th (midnight UCT) until November 8th (23:59 UTC)

See thread below for links are more info 👇

1/5

[Monocular Depth Estimation Challenge @DepthChallenge]

This year we’ll be hosting some great keynote talks by @ftm_guney, @eric_brachmann & @vitorguizilini! The organization team also includes 3 new members: Ripudaman Arora, @mattpoggi & @fabiotosi92 👋

For more details and updates, please check https://t.co/jGMTqM75oH

2/2

[Monocular Depth Estimation Challenge @DepthChallenge]

We’re very excited to announce that the 3rd edition of MDEC will be taking place at @CVPR in Seattle! The challenge, accepting both supervised and self-supervised approaches, will be running from 1st Feb - 25th Mar on CodaLab (https://t.co/yV3kq33vWH)

1/2

[Monocular Depth Estimation Challenge @DepthChallenge]

The 2nd edition of #MDEC at #CVPR2023 is right around the corner! Happening June 18th (08:30 - 12:00 PDT)

* In person: West 208 - 209

* Online: Zoom link at https://t.co/QXot8e9O4O

* Workshop details: https://t.co/jGMTqM7Def

@CVPR @cvssp_research

[Monocular Depth Estimation Challenge @DepthChallenge]

Less than a month left until #CVPR2023! 🇨🇦

We've got some great keynotes by @oisinmacaodha, Daniel Cremers and @alexgkendall, discussing monocular depth, 3D reconstruction and foundation models!

You can find more info at https://t.co/jGMTqM75oH

@cvssp_research @CVPR

[Monocular Depth Estimation Challenge @DepthChallenge]

Thank you to all participants! We are looking forward to seeing you all at the workshop! Details can be found in https://t.co/jGMTqM75oH

@CVPR #CVPR2023 @cvssp_research

4/4

[Monocular Depth Estimation Challenge @DepthChallenge]

The results and paper for the second edition of MDEC are out! You can find the paper at https://t.co/8PvdomvGTQ.

Congratulations to the winning teams DJI&ZJU (supervised) and imec-IDLab-UAntwerp (self-supervised) 🎉

1/4

[Monocular Depth Estimation Challenge @DepthChallenge]

Team imec-IDLab-UAntwerp instead relied on a ConvNeXt-v2-H encoder and an HR-Depth decoder (modified to use deformable convolutions). The network was trained using the min reconstruction photometric loss and edge-aware smoothness on the Kitti Eigen Benchmark dataset.

3/4

[Monocular Depth Estimation Challenge @DepthChallenge]

The 2nd #MDEC is now over! Thanks to all participants. Results will be released in the coming weeks.

We are also very excited to announce that the final workshop keynote will be given by Daniel Cremers! Stayed tuned for more details.

#CVPR2023 @cvssp_research

[Monocular Depth Estimation Challenge @DepthChallenge]

https://t.co/KkmxONa3y1

[Monocular Depth Estimation Challenge @DepthChallenge]

The recordings for the 1st edition of MDEC are now available at https://t.co/HAb8d4qJ47

You can also check them out in the thread below 👇

[Monocular Depth Estimation Challenge @DepthChallenge]

https://t.co/uMAt6ClpT0

[Monocular Depth Estimation Challenge @DepthChallenge]

Excited to announce that the 2nd Monocular Depth Estimation Challenge will be happening at #CVPR2023! Featuring:
🏆 New challenge accepting *any* type of supervision
🎙️Keynote talks by @alexgkendall and @oisinmacaodha

For more details check https://t.co/jGMTqM7Def

[Monocular Depth Estimation Challenge @DepthChallenge]

Reminder that the Monocular Depth Estimation Challenge Workshop for #WACV2023 will be happening tomorrow (Sat 7th)!

The workshop will be both in-person and virtual, starting at 08:20 HST (18:20 GMT).

Please find the details for the workshop at https://t.co/jGMTqM75oH

[Monocular Depth Estimation Challenge @DepthChallenge]

MonoViT

Based on the MonoViT architecture (https://t.co/PSkf5e2AMy), consisting of convolutional and MPViT encoder blocks. Trained using the min photometric reconstruction loss and proxy depth maps from a pretrained self-supervised stereo network.

[Monocular Depth Estimation Challenge @DepthChallenge]

The results for the first edition of MDEC can be found in https://t.co/Mt8zqJBXvI

Thank you to all participants! We hope to see you again in future editions

For additional details see below 👇

[Monocular Depth Estimation Challenge @DepthChallenge]

z.suri

The network used was ConvNeXt+DiffNet, trained with an additional stereo pose regression loss.