Skew Class-Balanced Re-Weighting for Unbiased Scene Graph Generation
An unbiased scene graph generation (SGG) algorithm referred to as Skew Class-Balanced Re-Weighting (SCR) is proposed for considering the unbiased predicate prediction caused by the long-tailed distribution. The prior works focus mainly on alleviating the deteriorating performances of the minority pr...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-03-01
|
| Series: | Machine Learning and Knowledge Extraction |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2504-4990/5/1/18 |
| _version_ | 1797610483370426368 |
|---|---|
| author | Haeyong Kang Chang D. Yoo |
| author_facet | Haeyong Kang Chang D. Yoo |
| author_sort | Haeyong Kang |
| collection | DOAJ |
| description | An unbiased scene graph generation (SGG) algorithm referred to as Skew Class-Balanced Re-Weighting (SCR) is proposed for considering the unbiased predicate prediction caused by the long-tailed distribution. The prior works focus mainly on alleviating the deteriorating performances of the minority predicate predictions, showing drastic dropping recall scores, i.e., losing the majority predicate performances. It has not yet correctly analyzed the trade-off between majority and minority predicate performances in the limited SGG datasets. In this paper, to alleviate the issue, the <i>Skew Class-Balanced Re-Weighting</i> (SCR) loss function is considered for the unbiased SGG models. Leveraged by the skewness of biased predicate predictions, the SCR estimates the target predicate weight coefficient and then re-weights more to the biased predicates for better trading-off between the majority predicates and the minority ones. Extensive experiments conducted on the standard Visual Genome dataset and Open Image V4 and V6 show the performances and generality of the SCR with the traditional SGG models. |
| first_indexed | 2024-03-11T06:15:52Z |
| format | Article |
| id | doaj.art-04fa41b4de374fd2a1a04c38693e339d |
| institution | Directory Open Access Journal |
| issn | 2504-4990 |
| language | English |
| last_indexed | 2024-03-11T06:15:52Z |
| publishDate | 2023-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Machine Learning and Knowledge Extraction |
| spelling | doaj.art-04fa41b4de374fd2a1a04c38693e339d2023-11-17T12:16:57ZengMDPI AGMachine Learning and Knowledge Extraction2504-49902023-03-015128730310.3390/make5010018Skew Class-Balanced Re-Weighting for Unbiased Scene Graph GenerationHaeyong Kang0Chang D. Yoo1School of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon 34141, Republic of KoreaSchool of Electrical Engineering, Korea Advanced Institute of Science and Technology (KAIST), Daejeon 34141, Republic of KoreaAn unbiased scene graph generation (SGG) algorithm referred to as Skew Class-Balanced Re-Weighting (SCR) is proposed for considering the unbiased predicate prediction caused by the long-tailed distribution. The prior works focus mainly on alleviating the deteriorating performances of the minority predicate predictions, showing drastic dropping recall scores, i.e., losing the majority predicate performances. It has not yet correctly analyzed the trade-off between majority and minority predicate performances in the limited SGG datasets. In this paper, to alleviate the issue, the <i>Skew Class-Balanced Re-Weighting</i> (SCR) loss function is considered for the unbiased SGG models. Leveraged by the skewness of biased predicate predictions, the SCR estimates the target predicate weight coefficient and then re-weights more to the biased predicates for better trading-off between the majority predicates and the minority ones. Extensive experiments conducted on the standard Visual Genome dataset and Open Image V4 and V6 show the performances and generality of the SCR with the traditional SGG models.https://www.mdpi.com/2504-4990/5/1/18scene graph generation (SGG)skew class-balanced re-weighting (SCR)predicate sample estimatesskew class-balanced effective number |
| spellingShingle | Haeyong Kang Chang D. Yoo Skew Class-Balanced Re-Weighting for Unbiased Scene Graph Generation Machine Learning and Knowledge Extraction scene graph generation (SGG) skew class-balanced re-weighting (SCR) predicate sample estimates skew class-balanced effective number |
| title | Skew Class-Balanced Re-Weighting for Unbiased Scene Graph Generation |
| title_full | Skew Class-Balanced Re-Weighting for Unbiased Scene Graph Generation |
| title_fullStr | Skew Class-Balanced Re-Weighting for Unbiased Scene Graph Generation |
| title_full_unstemmed | Skew Class-Balanced Re-Weighting for Unbiased Scene Graph Generation |
| title_short | Skew Class-Balanced Re-Weighting for Unbiased Scene Graph Generation |
| title_sort | skew class balanced re weighting for unbiased scene graph generation |
| topic | scene graph generation (SGG) skew class-balanced re-weighting (SCR) predicate sample estimates skew class-balanced effective number |
| url | https://www.mdpi.com/2504-4990/5/1/18 |
| work_keys_str_mv | AT haeyongkang skewclassbalancedreweightingforunbiasedscenegraphgeneration AT changdyoo skewclassbalancedreweightingforunbiasedscenegraphgeneration |