# karpathy/neuraltalk2

Efficient Image Captioning code in Torch, runs on GPU

Repository: https://github.com/karpathy/neuraltalk2
Canonical: https://ross.abutalabs.com/products/neuraltalk2
Language: Jupyter Notebook
License Family: other
Last push: 2017-11-07T05:59:20+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 3940, "days_push": 3221, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5592, forks 1267 (observed 2026-08-28T04:09:22.699492+00:00)

## What it is
NeuralTalk2 is a Torch (Lua) implementation of image captioning that uses a CNN (VGGNet) followed by an RNN language model to generate textual descriptions of images, with GPU-accelerated batched training and CNN finetuning support. It includes pretrained models and was a well-known research codebase, now superseded by TensorFlow's im2txt.

## Use cases
- generate captions for images with a neural network
- train an image captioning model on my own dataset
- evaluate image captioning models with CIDEr
- learn how CNN-RNN captioning models work
- run pretrained image captioning on GPU

## When to choose
- you specifically need a Torch/Lua image captioning implementation for education or research reproduction
- you want to study a classic CNN+RNN captioning architecture with readable inline comments

## When to avoid
- you need state-of-the-art captioning quality - use transformer-based or Google's im2txt models instead
- your stack is PyTorch or TensorFlow and you need maintained tooling
- you need a supported, actively maintained project

## Facets
- artifact type: library
- maturity: abandoned
- function: machine-learning, deep-learning, image-processing, nlp
- domain: machine-learning, computer-vision, deep-learning
- platform: python
- tags: image-captioning, torch, lstm, cnn, rnn, deprecated, natural-language-processing, linux, gpu

## Member repositories
- karpathy/neuraltalk2 (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:22.699492+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:55:30.046232+00:00, confidence not recorded.
  - readme: https://github.com/karpathy/neuraltalk2 (fetched 2026-08-28T04:09:22.699492+00:00, sha 673c7ca9c618)
- Data as of 2026-08-30T08:39:29.467469+00:00.