Corn pose estimation with 3D object detection

8/11 Abstract

8/16 Finish codes

8/20 Results

9/11 Dratf Finished

9/20 Images complated

9/24 V1 finished from Pro.Zhang

9/30 Grammar correction proposed

1. I have finished the first half of the new paper.

1.change the backbone of FCOS-STEREO to DLA-34 ,and try to make more improvement.

2. Based on cross-attention and efficientformer . i propose a light transformer architecture regression model.

1. The code using the EfficientFormer to regress the 3D parameters of kitti has been finsined.
2. I am trying to introduce the stereo structure into the EfficientFormer https://github.com/snap-research/EfficientFormer.

1. Finish chapter 1 of thesis.
2. Posedetection of the corn are progressing.
3. try to use the effientformer to regress the 3D result

1. Modify YOLOv7 structure for KITTI datasets.
2. The code of YOLOv7 is studied, and the loder and processing of stereo-data are introduced into the model structure.

My research aims to study the application of deep learning technology in corn cultivation to meet various needs at different stages of corn planting. In the stage of corn breeding, it is crucial to assess the growth status of corn, with phenotype extraction algorithms being key technologies. During the development and maturity phases … Continue Reading ››

1. Article has been published( https://doi.org/10.3390/agriculture14060907)
2. Describe the experience briefly:

   • The article should make it clear what purpose is to be accomplished and what results are to be achieved. What kind of methods were used? I think this is most improtant.

1. Paper access: ''Extraction of Corn Plant Phenotypic Parameters with Keypoint Detection and Stereo Images''(https://doi.org/10.3390/agronomy14061110)
2. Making research on 6D Pose and choosing a suitable frame to reproduce.

1.learn the framework of yolo-v8

2. learn the seam attention mechanism（https://arxiv.org/abs/2208.02019）

3. Try to come up with a better data fusion framework

4.Evaluation mechanisms are learning