CNF:ML Data Denoise - Revision history

Pthomadakis: /* Results */

2020-12-05T01:51:45Z

Results

← Older revision		Revision as of 21:51, 4 December 2020
Line 253:		Line 253:
	! Validation loss		! Validation loss
	! Training Time		! Training Time
	! Inference ! Train/Val Loss Plot		! Inference
			! Train/Val Loss Plot
	\|-		\|-
	\| 10		\| 10

Pthomadakis: /* Results */

2020-12-05T01:45:44Z

Results

← Older revision		Revision as of 21:45, 4 December 2020
Line 206:		Line 206:

	==== Results ====		==== Results ====
			{\|
			\|
			{\| class="wikitable
			! Noisy
			! Correct
			! Denoised
			\|-
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex1.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex1_cor.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex1_denoise.png]]
			\|-
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex2.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex2_cor.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex2_denoise.png]]
			\|-
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex3.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex3_cor.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex3_denoise.png]]
			\|}
			\|
			{\| class="wikitable
			! Noisy
			! Correct
			! Denoised
			\|-
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex4.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex4_cor.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex4_denoise.png]]
			\|-
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex5.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex5_cor.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex5_denoise.png]]
			\|-
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex6.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex6_cor.png]]
			\|[[File:cnf_ml_denoise_3d__250_24p_noisy2_ex6_denoise.png]]
			\|}
			\|}

			[[File:cnf_ml_denoise_3d__250_24p_noisy2_hits_hist.png]]
			[[File:cnf_ml_denoise_3d__250_24p_noisy2_noise_hist.png]]

			{\| class="wikitable
			! Epochs
			! Training loss
			! Validation loss
			! Training Time
			! Inference ! Train/Val Loss Plot
			\|-
			\| 10
			\| 0.021
			\| 0.022
			\| 1784
			\| 0.003
			\| [[File:cnf_ml_cnn_v2_train_val_loss_3d_250_24p.png\|480px]]
			\|}

	== 2D Input ==		== 2D Input ==

Pthomadakis: /* 250 Tracks */

2020-12-05T01:42:42Z

250 Tracks

← Older revision		Revision as of 21:42, 4 December 2020
Line 101:		Line 101:


	~~==== Data Augmentation ====~~
	~~The given dataset for 250 tracks only consists of 1000 data points. This hurts the ability of the network to learn.~~
	~~To overcome this limitation we augment the dataset with synthetic data points by using random combinations of the 1 track dataset tracks.~~
	{\|		{\|
	\| style="width: 20px" \| Example noisy input:		\| style="width: 20px" \| Example noisy input:
Line 111:		Line 109:
	\| [[File:cnf_ml_denoise_3d_250_2.png]]		\| [[File:cnf_ml_denoise_3d_250_2.png]]
	\|}		\|}
			==== Data Augmentation ====
			The given dataset for 250 tracks only consists of 1000 data points. This hurts the ability of the network to learn.
			To overcome this limitation we augment the dataset with synthetic data points by using random combinations of the 1 track dataset tracks.
	{\|		{\|
	\|		\|

Pthomadakis: /* 250 Tracks */

2020-12-05T01:42:28Z

250 Tracks

← Older revision		Revision as of 21:42, 4 December 2020
Line 99:		Line 99:

	=== 250 Tracks ===		=== 250 Tracks ===


			==== Data Augmentation ====
			The given dataset for 250 tracks only consists of 1000 data points. This hurts the ability of the network to learn.
			To overcome this limitation we augment the dataset with synthetic data points by using random combinations of the 1 track dataset tracks.
	{\|		{\|
	\| style="width: 20px" \| Example noisy input:		\| style="width: 20px" \| Example noisy input:
Line 164:		Line 169:
	\| [[File:cnf_ml_cnn_v2_train_val_loss_3d_250.png\|480px]]		\| [[File:cnf_ml_cnn_v2_train_val_loss_3d_250.png\|480px]]
	\|}		\|}

	~~==== Data Augmentation ====~~
	~~The given dataset for 250 tracks only consists of 1000 data points. This hurts the ability of the network to learn.~~
	~~To overcome this limitation we augment the dataset with synthetic data points by using random combinations of the 1 track dataset tracks.~~

	=== Z Plane Quantization ===		=== Z Plane Quantization ===

Pthomadakis: /* Z Plane Quantization */

2020-12-05T01:41:38Z

Z Plane Quantization

← Older revision		Revision as of 21:41, 4 December 2020
Line 203:		Line 203:
	\| style="text-align: center; vertical-align:top;"\| [[File:CNF_3D_QUANTIZATION_43.png]]		\| style="text-align: center; vertical-align:top;"\| [[File:CNF_3D_QUANTIZATION_43.png]]
	\|}		\|}

			==== Results ====

	== 2D Input ==		== 2D Input ==

Pthomadakis: /* 250 Tracks */

2020-12-04T22:04:26Z

250 Tracks

@@ Line 148: / Line 148: @@
 [[File:cnf_ml_denoise_3d__250_noisy2_hits_hist.png]]
 [[File:cnf_ml_denoise_3d__250_noisy2_noise_hist.png]]
 ==== Data Augmentation ====

Pthomadakis: /* 250 Tracks */

2020-12-04T21:53:00Z

250 Tracks

@@ Line 99: / Line 99: @@
 === 250 Tracks ===
 ==== Data Augmentation ====
 The given dataset for 250 tracks only consists of 1000 data points. This hurts the ability of the network to learn.

Pthomadakis: /* Data Augmentation */

2020-12-04T21:50:59Z

Data Augmentation

← Older revision		Revision as of 17:50, 4 December 2020
Line 101:		Line 101:
	==== Data Augmentation ====		==== Data Augmentation ====
	The given dataset for 250 tracks only consists of 1000 data points. This hurts the ability of the network to learn.		The given dataset for 250 tracks only consists of 1000 data points. This hurts the ability of the network to learn.
	To overcome this limitation we augment the dataset with synthetic data points by using random combinations of the 1 track dataset.		To overcome this limitation we augment the dataset with synthetic data points by using random combinations of the 1 track dataset tracks.

	=== Z Plane Quantization ===		=== Z Plane Quantization ===

Pthomadakis: /* 250 Tracks */

2020-12-04T21:50:31Z

250 Tracks

← Older revision		Revision as of 17:50, 4 December 2020
Line 99:		Line 99:

	=== 250 Tracks ===		=== 250 Tracks ===
			==== Data Augmentation ====
			The given dataset for 250 tracks only consists of 1000 data points. This hurts the ability of the network to learn.
			To overcome this limitation we augment the dataset with synthetic data points by using random combinations of the 1 track dataset.

	=== Z Plane Quantization ===		=== Z Plane Quantization ===

Pthomadakis: /* 3D Input */

2020-12-04T21:48:26Z

3D Input

← Older revision		Revision as of 17:48, 4 December 2020
Line 98:		Line 98:
	[[File:cnf_ml_denoise_3d_noisy2_noise_hist.png]]		[[File:cnf_ml_denoise_3d_noisy2_noise_hist.png]]

			=== 250 Tracks ===

	=== Z Plane Quantization ===		=== Z Plane Quantization ===