CORe50: a New Dataset and Benchmark for Continuous Object Recognition
About
Continuous/Lifelong learning of high-dimensional data streams is a challenging research problem. In fact, fully retraining models each time new data become available is infeasible, due to computational and storage issues, while na\"ive incremental strategies have been shown to suffer from catastrophic forgetting. In the context of real-world object recognition applications (e.g., robotic vision), where continuous learning is crucial, very few datasets and benchmarks are available to evaluate and compare emerging techniques. In this work we propose a new dataset and benchmark CORe50, specifically designed for continuous object recognition, and introduce baseline approaches for different continuous learning scenarios.
Vincenzo Lomonaco, Davide Maltoni• 2017
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Link Prediction | GraphHigher last snapshot (test) | MRR20.3 | 12 | |
| Link Prediction | GraphEqual last snapshot (test) | MRR12.2 | 12 | |
| Link Prediction | GraphLower last snapshot (test) | MRR3.2 | 12 | |
| Link Prediction | DB-CKGE last snapshot (test) | MRR0.026 | 12 | |
| Link Prediction | ENTITY (test) | MRR8.8 | 12 | |
| Link Prediction | RELATION (test) | MRR2.1 | 12 | |
| Link Prediction | Hybrid (test) | MRR0.037 | 12 | |
| Link Prediction | FB-CKGE (test) | MRR7.5 | 12 | |
| Object Classification | CORe-50 (test) | Increments21 | 10 | |
| Class-incremental learning | ImageNet first 10 classes 1.0 (test) | Accuracy (Task 1)83.2 | 4 |
Showing 10 of 10 rows