Comments (1)
hi, i want to konw how th AB is calculate? the code is as follows:
def send(self, key, val): self.data[key] = val if isinstance(val, np.ndarray): cur_bytes = val.size * 8 elif type(val) in [int, float]: cur_bytes = 8 elif isinstance(val, list): cur_bytes = np.array(val).size * 8 elif type(val) is str: cur_bytes = len(val) if key.endswith("boxes"): cur_bytes = cur_bytes * 7 / 24 self.cur_bytes += cur_bytesi want to konw which is used to calculate the point cloud or box?
for pointcloud, each point cloud has 4 float value, each float takes up 4 bytes. so the AB of pointcloud is N44 (N is the numbers of point clous in each file).
for box,each box has 8 vertices, each vertice has 3 coordinates. if it is float too, the AB of pointcloud is M83*4(M is the number of boxes in each file).
Am I doing the right thing?, Or can you tell me about the logic of the source code?
Thank you!
In our newer paper, we will normalize and explain how to calculate the transmission cost of the three transmission forms.
Here we use a 32-bit float to represent each element in transmitted data, and we ignore the transmission cost of calibration files and timestamps.
- Calculating the transmission cost of transmitting raw data for early fusion. We represent each point in point clouds with
$(x,y,z, intensity)$ , so each point requires four 32-bit floats that are equal to 16 Bytes. If there are 100,000 points in point clouds, the$\mathcal{AB}$ of the transmission cost is 1.6$\times 10^6$ Bytes. - Calculating the transmission cost of transmitting detection outputs for late fusion. We represent each 3D detection output with
$(x, y, z, w, l, h, \theta, confidence)$ , so each detection output requires eight 32-bit floats that are equal to 32 Bytes. If we transmit ten detection outputs per transmission on average, the$\mathcal{AB}$ of the transmission cost is 3.2$\times 10^2$ Bytes. - Calculating the transmission cost of transmitting intermediate data for middle fusion. The feature or feature flow is represented with the tensor, and we represent each element with a 32-bit float. If the size of the feature or feature flow is
$(100, 100, 100)$ , then the$\mathcal{AB}$ of the transmission cost is 100$\times$ 100$\times$ 100$\times$ 4 Bytes, which are equal to 4$\times 10^6$ Bytes.
from dair-v2x.
Related Issues (20)
- Dataset request -- V2X-Seq -- Outside Mainland China HOT 2
- Training HOT 3
- training HOT 2
- Frequently Asked & Good Questions
- DAIR-V2X-V 标签偏差大
- Different results on MMdetection3D and OpenPCDet for DAIR-V2X-I training
- Questions about the SPD installation process HOT 2
- ImportError: cannot import name 'DAIRV2XDataset' from partially initialized module 'base_dataset' (most likely due to a circular import) HOT 1
- 车路协同3D检测数据集 infrastructure-side 中image 的6143--6687缺失 HOT 5
- Hou to product the train.json,val.json from the data?
- Are the images in V2X-I dataset the subset of the images in Rope3D? HOT 1
- 路侧和车侧图像数据有对应序号吗
- 运行 bash scripts/eval_camera_late_fusion_imvoxelnet.sh 缺少包的问题
- Do you have any plan to publicly provide the dataset outside China? HOT 2
- early_fusion复现不出来
- 请问这些代码都只能在linux系统下运行对吗 HOT 1
- Question about 'compute_ap' function
- BEV Map Segmentation HOT 3
- KeyError: 'GroupFree3DMHA is already registered in attention'
- Error occurs when converting the dataset into Kitti format.
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 dair-v2x.