Dimensions overall score 7.0

GitHub Code Pulse

No public code linked for this paper yet.

• DEFOM-Stereo ViT-S achieves the best depth-domain accuracy on the synthetic test set with an EPE of 1.74 px and a depth MAE of 23.40 cm.result 95%
• DEFOM-Stereo ViT-S has an inference speed of approximately 2.2 FPS on the NVIDIA Jetson Orin Super 16 GB, which is too slow for responsive closed-loop tool control.limitation 95%
• DEFOM-PrunePlus offers the best deployable accuracy-speed trade-off, achieving approximately 3.3 FPS on the Jetson with an EPE of 5.87 px and a depth MAE of 64.26 cm.result 95%
• The synthetic dataset for training DEFOM-Stereo variants was built in Unreal Engine 5 using a simulated ZED Mini stereo camera, capturing 5,520 stereo pairs across 115 tree instances.method 95%
• DEFOM-PrunePlus provides a practical accuracy-latency balance for onboard distance estimation in autonomous UAV pruning.technical 90%
• Zero-shot inference on real photographs confirms that full-capacity DEFOM-Stereo models preserve branch geometry, validating the sim-to-real transfer.result 90%
• Lightweight variants DEFOM-PruneStereo and DEFOM-PruneNano achieve higher frame rates (6.9 FPS and 8.5 FPS respectively) but sacrifice substantial accuracy, with depth MAE greater than 57 cm.result 90%
• DEFOM-Stereo ViT-S achieves the best depth-domain accuracy on the synthetic test set with an EPE of 1.74 px and a depth MAE of 23.40 cm.result 95%
• DEFOM-Stereo ViT-S has an inference speed of approximately 2.2 FPS on the NVIDIA Jetson Orin Super 16 GB, which is too slow for responsive closed-loop tool control.limitation 95%
• DEFOM-PrunePlus offers the best deployable accuracy-speed trade-off with approximately 3.3 FPS on the Jetson and a depth MAE of 64.26 cm.result 95%
• The synthetic dataset for training DEFOM-Stereo variants was built in Unreal Engine 5 using a simulated ZED Mini stereo camera, capturing 5,520 stereo pairs across 115 tree instances.method 95%
• DEFOM-PrunePlus provides a practical accuracy-latency balance for onboard distance estimation for autonomous UAV pruning.technical 90%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card