ZhuoYi - Revision history

imported>Student221: /* Code */

2020-11-24T05:16:28Z

Code

← Older revision		Revision as of 05:16, 24 November 2020
Line 149:		Line 149:
	=== Code ===		=== Code ===

	* All code for this project can be downloaded here: [https://drive.google.com/file/d/1WdSpB0tZ8PZn0YN8KDrh435g9F6363VE/view?usp=sharing FinalProject.zip]		* All code for all experiments repro in this project can be downloaded here: [https://drive.google.com/file/d/1WdSpB0tZ8PZn0YN8KDrh435g9F6363VE/view?usp=sharing FinalProject.zip]

imported>Student221: /* Code */

2020-11-24T05:14:41Z

Code

← Older revision		Revision as of 05:14, 24 November 2020
Line 149:		Line 149:
	=== Code ===		=== Code ===

	* ~~Coming soon after presentation~~		* All code for this project can be downloaded here: [https://drive.google.com/file/d/1WdSpB0tZ8PZn0YN8KDrh435g9F6363VE/view?usp=sharing FinalProject.zip]

imported>Student221: /* Future Studies */

2020-11-24T05:08:18Z

Future Studies

← Older revision		Revision as of 05:08, 24 November 2020
Line 113:		Line 113:
	=== Future Studies ===		=== Future Studies ===

			* Why does compressed JPEG training 2D images result in better network performance and how come it makes the training process difficult to converge?
			** How does the network respond to the artifacts resulted from low quality JPEG images?
	* Why does compressed training 2D images result in better network performance and how come it makes the training process difficult to converge?

	* How reliable are the performance metrics?		* How reliable are the performance metrics?
			** Metrics may be misleading. What are the better metrics we can use for network performance under different applications?

	* For 2D image segmentation, why does Deeplab v3+ take longer time to train and have better performance when using 644x482 image dataset than even a higher resolution images? Does spatial aliasing play a role here?		* For 2D image segmentation, why does Deeplab v3+ take longer time to train and have better performance when using 644x482 image dataset than even a higher resolution images? Does spatial aliasing play a role here?

	* How ~~doest~~ gaussian noise on 2D image affect network performance?		* How does gaussian noise on 2D image affect network performance? Could this be related to the better network performance trained by the lower quality images we observed in this project?

	* Exploring how other networks behave with respect to training image qualities.		* Exploring how other networks behave with respect to training image qualities.

imported>Student221: /* Future Studies */

2020-11-24T05:02:37Z

Future Studies

← Older revision		Revision as of 05:02, 24 November 2020
Line 112:		Line 112:

	=== Future Studies ===		=== Future Studies ===



	* Why does compressed training 2D images result in better network performance and how come it makes the training process difficult to converge?		* Why does compressed training 2D images result in better network performance and how come it makes the training process difficult to converge?

imported>Student221: /* Observations */

2020-11-24T05:02:13Z

Observations

← Older revision		Revision as of 05:02, 24 November 2020
Line 107:		Line 107:
	* The point cloud segmentation network behaves differently from the 2D image segmentation network. Its performance is very sensitive to the volume and quality of the training data.		* The point cloud segmentation network behaves differently from the 2D image segmentation network. Its performance is very sensitive to the volume and quality of the training data.

	* The pixel imbalance of image classes affects the network performance in terms of segmentation accuracy. A weight may be applied to different classes to improve performance.		* The pixel imbalance of image classes affects the network performance in terms of segmentation accuracy. A weight may be applied to different classes accordingly to improve performance.

	* The network trained by lower resolution 2D images ~~results~~ in better performance.		* The network trained by lower resolution 2D images could result in better performance.

	=== Future Studies ===		=== Future Studies ===

imported>Student221: /* Observations */

2020-11-24T04:57:40Z

Observations

← Older revision		Revision as of 04:57, 24 November 2020
Line 106:		Line 106:

	* The point cloud segmentation network behaves differently from the 2D image segmentation network. Its performance is very sensitive to the volume and quality of the training data.		* The point cloud segmentation network behaves differently from the 2D image segmentation network. Its performance is very sensitive to the volume and quality of the training data.

			* The pixel imbalance of image classes affects the network performance in terms of segmentation accuracy. A weight may be applied to different classes to improve performance.

			* The network trained by lower resolution 2D images results in better performance.

	=== Future Studies ===		=== Future Studies ===

imported>Student221: /* Future Studies */

2020-11-20T06:02:41Z

Future Studies

← Older revision		Revision as of 06:02, 20 November 2020
Line 114:		Line 114:

	* For 2D image segmentation, why does Deeplab v3+ take longer time to train and have better performance when using 644x482 image dataset than even a higher resolution images? Does spatial aliasing play a role here?		* For 2D image segmentation, why does Deeplab v3+ take longer time to train and have better performance when using 644x482 image dataset than even a higher resolution images? Does spatial aliasing play a role here?

			* How doest gaussian noise on 2D image affect network performance?

			* Exploring how other networks behave with respect to training image qualities.

	== Acknowledgement==		== Acknowledgement==

imported>Student221: /* Correlation Analysis between Semantic Segmentation Performance and Data Quality */

2020-11-20T05:18:40Z

Correlation Analysis between Semantic Segmentation Performance and Data Quality

← Older revision		Revision as of 05:18, 20 November 2020
Line 1:		Line 1:
	== Correlation Analysis between Semantic Segmentation Performance and Data Quality==		== Correlation Analysis between Semantic Segmentation Performance and Training Data Quality==

	== Introduction ==		== Introduction ==

imported>Student221: /* Correlation Analysis between Semantic Segmentation Performance and Data Quality */

2020-11-20T04:34:18Z

Correlation Analysis between Semantic Segmentation Performance and Data Quality

← Older revision		Revision as of 04:34, 20 November 2020
Line 1:		Line 1:
	== Correlation Analysis between Semantic Segmentation Performance and Data Quality==		== Correlation Analysis between Semantic Segmentation Performance and Data Quality==

	~~Team: ZhuoYi Cai~~

	~~Term: Fall 2020~~

	== Introduction ==		== Introduction ==

imported>Student221: /* Future Studies */

2020-11-20T04:34:06Z

Future Studies

← Older revision		Revision as of 04:34, 20 November 2020
Line 117:		Line 117:
	* How reliable are the performance metrics?		* How reliable are the performance metrics?

	* For 2D image segmentation, why does Deeplab v3+ take longer time to train and have better performance when using 644x482 image dataset than even a higher resolution images?		* For 2D image segmentation, why does Deeplab v3+ take longer time to train and have better performance when using 644x482 image dataset than even a higher resolution images? Does spatial aliasing play a role here?

	== Acknowledgement==		== Acknowledgement==