OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Adabins

Adabins is a Python deep-learning implementation of adaptive-bin monocular depth estimation that predicts a depth map from an RGB image. It provides pretrained NYU and KITTI checkpoints and inference helpers for Pillow images, tensor batches and image directories.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Adabins — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Adabins when you need pretrained monocular depth predictions from RGB images and can work with its Python implementation and downloaded checkpoints.

Who it is for

  • Developers integrating single-image depth estimation into image-processing workflows.
  • Researchers or practitioners exploring pretrained depth prediction with NYU or KITTI checkpoints.

Documented uses

  • Generate a predicted depth map from one RGB image.
  • Run depth inference on a batch of RGB tensors.
  • Process a directory of images and save predicted depths in 16-bit format.
  • Choose an NYU or KITTI checkpoint with its corresponding example depth range.
Before you choose

Requirements & limits

  • The repository provides separate NYU and KITTI pretrained checkpoints; the example setup uses maximum depth values of 10 and 80 respectively.
  • The inference helper handles preprocessing, uses horizontal-flip test-time augmentation and calculates bin centers.
  • The project is identified as GPL-3.0 licensed.
  • The README documents inference examples, while evaluation and training instructions remain listed as TODOs.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Generate depth predictions with a pretrained checkpoint

Use the provided inference helper to predict depth for an image or a directory of images.

Before you start

  • The AdaBins repository and its Python code.
  • A downloaded pretrained checkpoint: AdaBins_nyu.pt or AdaBins_kitti.pt.
  • An RGB Pillow image, RGB tensor batch, or input directory containing only images.

Steps

  1. Download the NYU or KITTI pretrained model and place the checkpoint in a directory of your choice.

  2. Create an InferenceHelper configured for the corresponding dataset, such as InferenceHelper(dataset='nyu').

  3. Call predict_pil for a Pillow image, predict for an RGB tensor batch, or predict_dir with input and output directories.

Expected outcome

The helper returns bin centers and predicted depth for image or batch inference; directory inference stores predicted depths in 16-bit format in the specified output directory.

Guide limitations

  • The documented directory input should contain only images.
  • Evaluation and training instructions are listed as TODO items in the README.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
18 October 2026
Original website title
shariqfarooq123/AdaBins · Official homepage title · retained in its source language
How we review sourcesSuggest a correction