SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.16034 · ECCV · 2022

Monitored Distillation for Positive Congruent Depth Completion

Alex Wong, Tian Liu, Parth Agrawal, Allison Chen, Byung-Woo Hong

arXiv · PDF · Open in the Atlas

Code that ran

We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

We propose a method to infer a dense depth map from a single image, its calibration, and the associated sparse point cloud. In order to leverage existing models (teachers) that produce putative depth maps, we propose an adaptive knowledge distillation approach that yields a positive congruent training process, wherein a student model avoids learning the error modes of the teachers. In the absence of ground truth for model selection and training, our method, termed Monitored Distillation, allows a student to exploit a blind ensemble of teachers by selectively learning from predictions that best minimize the reconstruction error for a given image. Monitored Distillation yields a distilled depth map and a confidence map, or "monitor", for how well a prediction from a particular teacher fits the observed image. The monitor adaptively weights the distilled depth where if all of the teachers exhibit high residuals, the standard unsupervised image reconstruction loss takes over as the supervisory signal. On indoor scenes (VOID), we outperform blind ensembling baselines by 17.53% and unsupervised methods by 24.25%; we boast a 79% model size reduction while maintaining comparable performance to the best supervised method. For outdoors (KITTI), we tie for 5th overall on the benchmark despite not using ground truth. Code available at: https://github.com/alexklwong/mondi-python.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2203.16034")
get_code_for_paper("2203.16034")
have("2203.16034")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2203.16034.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2203.16034.svg)](https://syntology.ai/paper/2203.16034)

Connect an agent — have() is free.