search

tnikolla / robot-grasp-detection

Detecting robot grasping positions with deep neural networks. The model is trained on Cornell Grasping Dataset. This is an implementation mainly based on the paper 'Real-Time Grasp Detection Using Convolutional Neural Networks' from Redmon and Angelova.
Apache License 2.0
232 stars 84 forks source link

This network only works on the object of the image center? #13

Open HEUzhouhanwen opened 6 years ago

HEUzhouhanwen commented 6 years ago

I used other images to test it and found that this network only works on the object of the image center!!!

Why is it like this?

Thank you! Shu Guo

xiaoshuguo750 commented 6 years ago

figure_1 figure_1-1

xiaoshuguo750 commented 6 years ago

figure_1-2

ahundt commented 6 years ago

Could you guys make a PR with your code to visualize this or let me know the commands? I'm also trying to use the dataset

ahundt commented 6 years ago

To answer your question, what data augmentation are you doing? There will need to be random crops, rotations, etc for best results.

Also, why do those images look like parallelograms? Shouldn't they be rectangles?

oslo71 commented 4 years ago

@xiaoshuguo750 I am a green hand to Deep Learning. could you please show me the code of predicting? Here is my Wechat number: 15189826919. I am looking forward to your help.

lijialei666 commented 3 years ago

I used other images to test it and found that this network only works on the object of the image center!!!

Why is it like this?

Thank you! Shu Guo 你好，这个问题解决了吗？只能检测中心位置

lijialei666 commented 3 years ago

@xiaoshuguo750 你好，这个问题解决了吗？只能检测中心位置

lijialei666 commented 3 years ago

To answer your question, what data augmentation are you doing? There will need to be random crops, rotations, etc for best results.

Also, why do those images look like parallelograms? Shouldn't they be rectangles?

Hello, why can only the image center be detected the correct results? Have you solved this problem? Do you have the latest model weights? Thank you.

lijialei666 commented 3 years ago

@ahundt @xiaoshuguo750 @HEUzhouhanwen @oslo71 Hello, why can only the image center be detected the correct results? Have you solved this problem? Do you have the latest model weights? Thank you.

zhoumo1121 commented 2 years ago

Hello, may I ask how to solve this problem?(grasp_det.py) ValueError: Tensor conversion requested dtype int32 for Tensor with dtype float64: 'Tensor("truediv:0", shape=(), dtype=float64, device=/device:CPU:0)'

Jonho111 commented 2 years ago

I used other images to test it and found that this network only works on the object of the image center!!!

Why is it like this?

Thank you! Shu Guo

hello,bro，i am a new guy to learn the robotic grasp,may i ask how to predict single picture with this code?