Comments (7)
如果是推理的话,只需要提供输入图像、相机内参和相机外参即可,对应代码中的变量就是input, gt_intrinsic和gt_extrinsic。
cam_height和pitch是从外参中计算得到,不需要额外设定。修改org_w和org_h即可。但目前代码只适配了OpenLane的数据格式,支持其训练和测试,还不能直接用于其他数据集(需要对应适配DataLoader)。对于单张图像的推理,一个简单的方式是按照OpenLane的数据集格式和路径格式准备对应的图片和标注(相机内参和相机外参从标注中读取)作为validation。
此外,由于nuscenes没有车道线标注,所以无法进行训练和定量的测试对比。需要注意的是,不经过finetune,直接使用OpenLane的预训练模型可能会由于数据集分布的差异导致效果不佳。
from persformer_3dlane.
@dyfcalid 感谢解答。
我试着测试了,用你们的数据集正常;用nuscene时发现生成的vis_ipm图像不对,怀疑是json里extrinsic参数有问题。那么extrinsic这个参数,是指camera to ego的Transformation Matrix吗,或者是world to camera?
from persformer_3dlane.
@dyfcalid 感谢解答。 我试着测试了,用你们的数据集正常;用nuscene时发现生成的vis_ipm图像不对,怀疑是json里extrinsic参数有问题。那么extrinsic这个参数,是指camera to ego的Transformation Matrix吗,或者是world to camera?
是的,camera to ego,相机坐标系到车辆坐标系
from persformer_3dlane.
你好,直接使用carema to ego测试结果是正确的吗?因为我尝试了一下ipm的可视化结果有问题,是因为openlane标注的waymo数据集的x,y和z轴和正常的偏向不同导致吗?
from persformer_3dlane.
from persformer_3dlane.
I used the original code of the Persformer model to predict front-view images in the NuScenes dataset, and I used the corresponding intrinsic and extrinsic parameters. However, when using Persformer for inference, the generated JSON results in result_3d show empty lane_lines. Besides changing the intrinsic and extrinsic parameters in the ground truth file, are there any other parameters that need to be adjusted during testing?
from persformer_3dlane.
Perhaps it's an issue with my input coordinate system? Because, upon investigating the problem, I found that during the compute_3d_lanes_all_category process, all the lane lines are being categorized as invalid lane lines (if np.argmax(category) == 0: continue). Consequently, the output 3D points are empty. I wanted to ask, during the testing of the NuScenes dataset with the camera parameters from the dataset,and chaning the ori_img_shape, have there been any modifications?"
from persformer_3dlane.
Related Issues (20)
- Will apollo dataset pretrained model release?
- ZeroDivisionError: Weights sum to zero, can't be normalized HOT 2
- Some question about the results of x_far,x_close,z_far,z_close? HOT 1
- 加载EfficientNet模型报错“unexpected EOF”
- visualization HOT 2
- JSON file Error
- Single Inference HOT 1
- Unable to replicate the performance of GenLaneNet and 3DLaneNet on the openlane dataset based on your framework.
- sheared feature mapping
- TypeError: 'module' object is not callable HOT 2
- 分布训练
- 关于类别数问题 HOT 3
- 不使用分布式训练
- UnboundLocalError: local variable '_label_cam_height' referenced before assignment HOT 1
- apollo dataset pretrained model
- Some questions on code
- how to derive H_g2cam?
- 推理其他数据集发生错误
- Lane Detection Results
- nms installed error
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from persformer_3dlane.