Webly Supervised Concept Expansion for General Purpose Vision Models

By Amita Kamath, Christopher Clark, Tanmay Gupta, Eric Kolve, Derek Hoiem, Aniruddha Kembhavi. 2022

72 citations recorded by the source corpus.

General Purpose Vision (GPV) systems are models that are designed to solve a wide array of visual tasks without requiring architectural changes. Today, GPVs primarily learn both skills and concepts from large fully supervised datasets. Scaling GPVs to tens of thousands of concepts by acquiring data to learn each concept for every skill quickly becomes prohibitive. This work presents an effective and inexpensive alternative: learn skills from supervised datasets, learn concepts from web image search, and leverage a key characteristic of GPVs: the ability to transfer visual knowledge across skills. We use a dataset of 1M+ images spanning 10k+ visual concepts to demonstrate webly-supervised concept expansion for two existing GPVs (GPV-1 and VL-T5) on 3 benchmarks: 5 COCO-based datasets (80 primary concepts), a newly curated series of 5 datasets based on the OpenImages and VisualGenome repositories (~500 concepts), and the Web-derived dataset (10k+ concepts). We also propose a new architecture, GPV-2 that supports a variety of tasks -- from vision tasks like classification and localization to vision+language tasks like QA and captioning, to more niche ones like human-object interaction detection. GPV-2 benefits hugely from web data and outperforms GPV-1 and VL-T5 across these benchmarks. Our data, code, and web demo are available at https://prior.allenai.org/projects/gpv2.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Unified-IO: A Unified Model for Vision, Language, and Multi-Modal Tasks (2022, methodology): “The next best submission is GPV-2 (Kamath et al., 2022) which obtains 32.”
PromptCap: Prompt-Guided Task-Aware Image Captioning (2022, background): “KAT [18] and GPV-2 [23] are the state-of-the-art supervised models for OK-VQA and A-OKVQA, respectively.”
Satlas: A Large-Scale, Multi-Task Dataset for Remote Sensing Image Understanding (2022, methodology): “Thus, inspired by recent work that integrate task-speciﬁc output heads [20, 27, 32, 33], we develop a uniﬁed model called S AT N ET that incorporates three such heads so that it can learn from every category in the dataset.”
See, Think, Confirm: Interactive Prompting Between Vision and Language Models for Knowledge-based Visual Reasoning (2023, unknown): “We found that our model has an accuracy of 57.5 on the test split of A-OKVQA, achieving better performance than previous fully-supervised methods like GPV-2 [35].”
A-OKVQA: A Benchmark for Visual Question Answering using World Knowledge (2022, methodology): “The models we consider are Pythia [20], VilBERT [28], LXMERT [45], KRISP [31], and GPV-2 [22].”