Codesota · Benchmark · CTW1500Home/Leaderboards/Vision & Documents/Scene Text Detection/CTW1500
Unknown

CTW1500.

1500 images with curved text annotations. Focus on arbitrary-shaped text.

Paper ↗Leaderboard ↓
§ 01 · SOTA history

Year over year.

§ 02 · Leaderboard

Results by metric.

Found a wrong score or missing run?
Use row edits to send a sourced correction into moderation.
Add / edit result ↗Report issue ↗

precision

Precision is the reported evaluation metric for CTW1500. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for precisionverifiedpapervendorcommunityunverified
RankModelTrustScoreYearLinksFix
01DBNet++ (ResNet-50) (1024)
ResNet-50, 1024px input. Table III.
verified88.52022Paper ↗Looks wrong?
02DBNet++ (ResNet-50) (800)
ResNet-50, 800px input. Table III.
verified87.92022Paper ↗Looks wrong?
03TextFuseNet (ResNeXt-101)
IJCAI 2020.
verified87.82020Paper ↗Looks wrong?
04DBNet++ (ResNet-18) (1024)
ResNet-18, 1024px input. Table III.
verified86.72022Paper ↗Looks wrong?
05CRAFT
Table 2 in CRAFT paper (CVPR 2019).
verified862019Paper ↗Looks wrong?
06DBNet++ (ResNet-18) (800)
ResNet-18, 800px input. Table III.
verified84.32022Paper ↗Looks wrong?

F Measure

F Measure is the reported evaluation metric for CTW1500. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for F Measureverifiedpapervendorcommunityunverified
RankModelTrustScoreYearLinksFix
01TextFuseNet (ResNeXt-101)
ResNet-101 backbone. P=87.8, R=85.4. IJCAI 2020.
verified86.62020Paper ↗Looks wrong?
02DBNet++ (ResNet-50) (800)
ResNet-50, 800px input, FPS=26. Table III.
verified85.32022Paper ↗Looks wrong?
03DBNet++ (ResNet-50) (1024)
ResNet-50, 1024px input, FPS=21. Table III.
verified85.12022Paper ↗Looks wrong?
04DBNet++ (ResNet-18) (1024)
ResNet-18, 1024px input, FPS=40. Table III.
verified83.92022Paper ↗Looks wrong?
05CRAFT
Table 2 in CRAFT paper (CVPR 2019).
verified83.52019Paper ↗Looks wrong?
06DBNet++ (ResNet-18) (800)
ResNet-18, 800px input, FPS=49. Table III.
verified82.62022Paper ↗Looks wrong?

recall

Recall is the reported evaluation metric for CTW1500. Codesota tracks published model scores on this metric so readers can compare state-of-the-art results across sources and model families.

Higher is better

Trust tiers for recallverifiedpapervendorcommunityunverified
RankModelTrustScoreYearLinksFix
01TextFuseNet (ResNeXt-101)
IJCAI 2020.
verified85.42020Paper ↗Looks wrong?
02DBNet++ (ResNet-50) (800)
ResNet-50, 800px input. Table III.
verified82.82022Paper ↗Looks wrong?
03DBNet++ (ResNet-50) (1024)
ResNet-50, 1024px input. Table III.
verified822022Paper ↗Looks wrong?
04DBNet++ (ResNet-18) (1024)
ResNet-18, 1024px input. Table III.
verified81.32022Paper ↗Looks wrong?
05CRAFT
Table 2 in CRAFT paper (CVPR 2019).
verified81.12019Paper ↗Looks wrong?
06DBNet++ (ResNet-18) (800)
ResNet-18, 800px input. Table III.
verified812022Paper ↗Looks wrong?
§ 04 · Submit a result

Add to the leaderboard.

← Back to Scene Text Detection