Finding Four-Leaf Clovers: A Benchmark for Fine-Grained ...

Post on 05-Jan-2022

3 views 0 download

Transcript of Finding Four-Leaf Clovers: A Benchmark for Fine-Grained ...

Finding Four-Leaf Clovers:A Benchmark for Fine-Grained Object Localization

Gustavo Pérez*, Laura Bravo*, Alejandro Pardo*, Pablo Arbeláez

01/15

*Indicates equal contribution

Finding Four-Leaf Clovers

Goal: to create a reliable benchmark for fine-grained object localization problems

Other fine-grained object localization problems:

Cancer diagnosis (Medicine)

Identifying cells in mitosis (Medicine)

Detecting infected crops (Biology)

Finding a specific person in a crowd (Industry)

Available at: tinyurl.com/yy9rbbsu 02/15

Finding Four-Leaf Clovers

High intra-class variability Low inter-class variability

Available at: tinyurl.com/yy9rbbsu 03/15

Finding Four-Leaf Clovers

Highly un-balanced

Available at: tinyurl.com/yy9rbbsu 04/15

Finding Four-Leaf Clovers

Available at: tinyurl.com/yy9rbbsu 05/15

Finding Four-Leaf Clovers

Available at: tinyurl.com/yy9rbbsu 06/15

The Four-Leaf Clover Dataset

Examples of the level of detail in segmentation annotations of the FLC dataset

Available at: tinyurl.com/yy9rbbsu 07/15

The Four-Leaf Clover Dataset

FLC dataset statistics. 4-leaf clover pixels and 4-leaf clover boundary pixels refer to the rate of the total of positive pixels over the total of pixels in the FLC dataset.

Available at: tinyurl.com/yy9rbbsu 08/15

Challenges of the FLC Dataset

Available at: tinyurl.com/yy9rbbsu 09/15

Tasks

Available at: tinyurl.com/yy9rbbsu 10/15

Experiments

Available at: tinyurl.com/yy9rbbsu 11/15

Experiments

Available at: tinyurl.com/yy9rbbsu 12/15

Experiments

Available at: tinyurl.com/yy9rbbsu 13/15

Experiments

Available at: tinyurl.com/yy9rbbsu 14/15

Thank you!

Available at: tinyurl.com/yy9rbbsu 15/15

8-leaf clover →In case you wonder...

Comparison to Other Datasets

Comparison of FLC to major visual recognition datasets. Club (♣) indicates that a dataset allows to study a recognition problem at a fine-grained level, triangle (Δ) indicates that the version of

the problem is not fine-grained, and (×) indicates that a dataset does not allow to study a problem.

Available at: tinyurl.com/yy9rbbsu