YueYawenXianzhe - Revision history

imported>Student221: /* Conclusion */

2019-12-14T07:35:23Z

Conclusion

← Older revision		Revision as of 07:35, 14 December 2019
Line 132:		Line 132:

	== Conclusion ==		== Conclusion ==
	The ~~result shows~~ a promising result and improvement with data augmentation. The RefineNet could improve the precision greatly based on the result of the CoarseNet. Currently, the design of RefineNet is relatively simple comparing to the CoarseNet, further tuning and modification on RefineNet are expected.		The evaluation and visualization show a promising result and improvement with data augmentation. The RefineNet could improve the precision greatly based on the result of the CoarseNet. Currently, the design of RefineNet is relatively simple comparing to the CoarseNet, further tuning and modification on RefineNet are expected.

	But there are many aspects we could improve in the future.		But there are many aspects we could improve in the future. We might need to train our model on a larger dataset first and then train it on this synthesized dataset. We also want to train our model on datasets in different domains to make it more robust to avoid errors introduced by this specific domain.

	== References ==		== References ==

imported>Student221: /* Conclusion */

2019-12-14T07:31:58Z

Conclusion

← Older revision		Revision as of 07:31, 14 December 2019
Line 134:		Line 134:
	The result shows a promising result and improvement with data augmentation. The RefineNet could improve the precision greatly based on the result of the CoarseNet. Currently, the design of RefineNet is relatively simple comparing to the CoarseNet, further tuning and modification on RefineNet are expected.		The result shows a promising result and improvement with data augmentation. The RefineNet could improve the precision greatly based on the result of the CoarseNet. Currently, the design of RefineNet is relatively simple comparing to the CoarseNet, further tuning and modification on RefineNet are expected.

	But there are many aspects we could improve in the future.		But there are many aspects we could improve in the future.

	== References ==		== References ==

imported>Student221: /* Conclusion */

2019-12-14T07:31:45Z

Conclusion

← Older revision		Revision as of 07:31, 14 December 2019
Line 132:		Line 132:

	== Conclusion ==		== Conclusion ==
	The result shows a promising result and improvement with data augmentation. But there are many aspects we could improve in the future.		The result shows a promising result and improvement with data augmentation. The RefineNet could improve the precision greatly based on the result of the CoarseNet. Currently, the design of RefineNet is relatively simple comparing to the CoarseNet, further tuning and modification on RefineNet are expected.

			But there are many aspects we could improve in the future.

	== References ==		== References ==

imported>Student221: /* Results */

2019-12-14T07:27:44Z

Results

← Older revision		Revision as of 07:27, 14 December 2019
Line 83:		Line 83:

	== Results ==		== Results ==
			=== Evaluation ===
	{\| class="Evaluation" width="auto" style="margin-left: auto; margin-right: auto; border: none;"		{\| class="Evaluation" width="auto" style="margin-left: auto; margin-right: auto; border: none;"
	\|-		\|-

imported>Student221: /* Analysis */

2019-12-14T07:25:20Z

Analysis

← Older revision		Revision as of 07:25, 14 December 2019
Line 126:		Line 126:

	=== Analysis ===		=== Analysis ===
	From the result above, we could observe the improvement with augmented dataset over the original dataset and our model gives us an acceptable depth map prediction precision.		From the result above, we could observe the improvement with augmented dataset over the original dataset and our model gives us an acceptable depth map prediction precision. For linear RMSE, the data augmentation gives us an improvement of 19.8% on CoarseNet prediction and 10.2% on RefineNet prediction. From the predicted depth map, we could get the rough depth information about the image. The error images show that we lose some details.
	we can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has fewer errors when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.
			However, there are some issues we might need to pay attention to. We can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has fewer errors when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.

	== Conclusion ==		== Conclusion ==

imported>Student221: /* Analysis */

2019-12-14T07:19:03Z

Analysis

← Older revision		Revision as of 07:19, 14 December 2019
Line 126:		Line 126:

	=== Analysis ===		=== Analysis ===
	From the result above, we can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has fewer errors when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.		From the result above, we could observe the improvement with augmented dataset over the original dataset and our model gives us an acceptable depth map prediction precision.
			we can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has fewer errors when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.

	== Conclusion ==		== Conclusion ==

imported>Student221: /* Conclusion */

2019-12-14T07:16:55Z

Conclusion

← Older revision		Revision as of 07:16, 14 December 2019
Line 129:		Line 129:

	== Conclusion ==		== Conclusion ==
			The result shows a promising result and improvement with data augmentation. But there are many aspects we could improve in the future.

	== References ==		== References ==

imported>Student221 at 07:13, 14 December 2019

2019-12-14T07:13:30Z

← Older revision		Revision as of 07:13, 14 December 2019
Line 125:		Line 125:
	[[File:Data_20.png\|1000px]]		[[File:Data_20.png\|1000px]]

	== Analysis ==		=== Analysis ===
	From the result above, we can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has fewer errors when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.		From the result above, we can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has fewer errors when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.

imported>Student221 at 07:12, 14 December 2019

2019-12-14T07:12:53Z

← Older revision		Revision as of 07:12, 14 December 2019
Line 126:		Line 126:

	== Analysis ==		== Analysis ==
	From the result above, we can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has ~~less error~~ when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.		From the result above, we can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has fewer errors when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.

			== Conclusion ==


	== References ==		== References ==

imported>Student221: /* Conclusions */

2019-12-14T07:11:54Z

Conclusions

← Older revision		Revision as of 07:11, 14 December 2019
Line 125:		Line 125:
	[[File:Data_20.png\|1000px]]		[[File:Data_20.png\|1000px]]

	== ~~Conclusions~~ ==		== Analysis ==
	From the result above, we can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has less error when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.		From the result above, we can find that our model learned limited information from the input data and there are more aspects to improve. Especially, it seems that our model tends to learn more about the intensity of the input image instead of the disparity information. The output of our model has less error when it estimates the disparity of a test image with a bright sky. However, since our input data contains a lot of street images with a sky, it tends to predict all brighter input with larger depth, just like the sky. This suggests that with a dataset of relatively similar scenes, our model tends to learn the characteristics of the object itself, such as intensity, instead of its disparity. Moreover, to improve our work, we can continue to refine the architecture of the mode, perform more data augmentation, tune hyperparameters, and test on other datasets for more precise depth estimation.