DepthCues: a bench to evaluate monocular depth perception of large vision models via linear probe.
It consists of 6 depth related tasks to assess response to visual cues.
They test 20 pretrained models.
As expected DAv2, DUSt3R, DINOv2 do well, but SigLIP is not bad
danier97.github.io/depthcues/
It consists of 6 depth related tasks to assess response to visual cues.
They test 20 pretrained models.
As expected DAv2, DUSt3R, DINOv2 do well, but SigLIP is not bad
danier97.github.io/depthcues/
November 28, 2024 at 12:12 AM
DepthCues: a bench to evaluate monocular depth perception of large vision models via linear probe.
It consists of 6 depth related tasks to assess response to visual cues.
They test 20 pretrained models.
As expected DAv2, DUSt3R, DINOv2 do well, but SigLIP is not bad
danier97.github.io/depthcues/
It consists of 6 depth related tasks to assess response to visual cues.
They test 20 pretrained models.
As expected DAv2, DUSt3R, DINOv2 do well, but SigLIP is not bad
danier97.github.io/depthcues/
The Blue Ridge Mountains along the Parkway! #MountainMonday #BlueSkyMonday #EastCoastKin #Knowledge #SensoryArt As a vision scientist, I love the shading of green to blue helping is appreciate the depth. #AtmosphericPerspective #depthcues #VisualPerception
September 15, 2025 at 2:41 PM
The Blue Ridge Mountains along the Parkway! #MountainMonday #BlueSkyMonday #EastCoastKin #Knowledge #SensoryArt As a vision scientist, I love the shading of green to blue helping is appreciate the depth. #AtmosphericPerspective #depthcues #VisualPerception
Along the Blue Ridge Parkway for #ForestFriday showing all the hues of the forest as it recedes into the distance. Taken several years ago. #EastCoastKin #scape #landscape #BlueRidgeParkway #AtmosphericPerspective #perspective #SensoryArt #NorthCarolina #SensationAndPerception #depthcues
June 27, 2025 at 11:34 AM
Along the Blue Ridge Parkway for #ForestFriday showing all the hues of the forest as it recedes into the distance. Taken several years ago. #EastCoastKin #scape #landscape #BlueRidgeParkway #AtmosphericPerspective #perspective #SensoryArt #NorthCarolina #SensationAndPerception #depthcues
We had a debate one year about how big of a tree we should get. A little advent silliness. #ArtAdventCalendar #EastCoastKin #Christmas #ChristmasTree #HolidaySeason #ForcedPerspective #SensationAndPerception #depthcues
December 11, 2025 at 1:19 PM
We had a debate one year about how big of a tree we should get. A little advent silliness. #ArtAdventCalendar #EastCoastKin #Christmas #ChristmasTree #HolidaySeason #ForcedPerspective #SensationAndPerception #depthcues
Here is a wonderful use of Atmospheric Perspective to enhance the perception of depth and distance.
#ColorADay #BlueTue #GreatSmokyMountains #AtmosphericPerspective #DepthCues #Knowledge #SensoryArt #EastCoastKin
#ColorADay #BlueTue #GreatSmokyMountains #AtmosphericPerspective #DepthCues #Knowledge #SensoryArt #EastCoastKin
Here's one. I'm in Tennessee, and this is based on a photo that a co-worker took while on vacation in the Great Smoky Mountains, several hours east of me.
September 16, 2025 at 4:59 PM
Here is a wonderful use of Atmospheric Perspective to enhance the perception of depth and distance.
#ColorADay #BlueTue #GreatSmokyMountains #AtmosphericPerspective #DepthCues #Knowledge #SensoryArt #EastCoastKin
#ColorADay #BlueTue #GreatSmokyMountains #AtmosphericPerspective #DepthCues #Knowledge #SensoryArt #EastCoastKin
DepthCues: Evaluating Monocular Depth Perception in Large Vision Models
Do automated monocular depth estimation methods use similar visual cues to humans?
To learn more, stop by poster #405 in the evening session (17:00 to 19:00) today at #CVPR2025.
Do automated monocular depth estimation methods use similar visual cues to humans?
To learn more, stop by poster #405 in the evening session (17:00 to 19:00) today at #CVPR2025.
June 14, 2025 at 12:48 PM
DepthCues: Evaluating Monocular Depth Perception in Large Vision Models
Do automated monocular depth estimation methods use similar visual cues to humans?
To learn more, stop by poster #405 in the evening session (17:00 to 19:00) today at #CVPR2025.
Do automated monocular depth estimation methods use similar visual cues to humans?
To learn more, stop by poster #405 in the evening session (17:00 to 19:00) today at #CVPR2025.
Look at the atmospheric perspective here! #scape #EastCoastKin #landscape #depthcues #perception #vision
February 14, 2025 at 4:56 PM
Look at the atmospheric perspective here! #scape #EastCoastKin #landscape #depthcues #perception #vision
Interesting experiments! We explored similar cues (occlusion, horizon, elevation, etc.), albeit in real-world images, across various vision models (including mono-depth) in our recent paper: arxiv.org/abs/2411.17385
DepthCues: Evaluating Monocular Depth Perception in Large Vision Models
Large-scale pre-trained vision models are becoming increasingly prevalent, offering expressive and generalizable visual representations that benefit various downstream tasks. Recent studies on the eme...
arxiv.org
February 17, 2025 at 6:12 PM
Interesting experiments! We explored similar cues (occlusion, horizon, elevation, etc.), albeit in real-world images, across various vision models (including mono-depth) in our recent paper: arxiv.org/abs/2411.17385
Duolikun Danier, Mehmet Ayg\"un, Changjian Li, Hakan Bilen, Oisin Mac Aodha
DepthCues: Evaluating Monocular Depth Perception in Large Vision Models
https://arxiv.org/abs/2411.17385
DepthCues: Evaluating Monocular Depth Perception in Large Vision Models
https://arxiv.org/abs/2411.17385
November 27, 2024 at 6:31 PM
Duolikun Danier, Mehmet Ayg\"un, Changjian Li, Hakan Bilen, Oisin Mac Aodha
DepthCues: Evaluating Monocular Depth Perception in Large Vision Models
https://arxiv.org/abs/2411.17385
DepthCues: Evaluating Monocular Depth Perception in Large Vision Models
https://arxiv.org/abs/2411.17385
Paper:
arxiv.org/abs/2411.17385
Project page:
danier97.github.io/depthcues/
Work led by Duolikun Danier:
danier97.github.io
arxiv.org/abs/2411.17385
Project page:
danier97.github.io/depthcues/
Work led by Duolikun Danier:
danier97.github.io
June 14, 2025 at 12:48 PM
Paper:
arxiv.org/abs/2411.17385
Project page:
danier97.github.io/depthcues/
Work led by Duolikun Danier:
danier97.github.io
arxiv.org/abs/2411.17385
Project page:
danier97.github.io/depthcues/
Work led by Duolikun Danier:
danier97.github.io
To explore this question, we developed a new benchmark, DepthCues, to evaluate human-like monocular depth cues in large vision models. We show that strong performance on these cues emerges in more recent larger monocular depth models.
June 14, 2025 at 12:48 PM
To explore this question, we developed a new benchmark, DepthCues, to evaluate human-like monocular depth cues in large vision models. We show that strong performance on these cues emerges in more recent larger monocular depth models.