ML-mini-project - Model testing

Mathilde Bindslev / ML-mini-project Public

Target: Nvidia Jetson Nano Clone this project

This lists all test data. You can manage this data through Data acquisition.

Test data

Set the 'expected outcome' for each sample to the desired outcome to automatically score the impulse.

Sample name	Expected outcome	Length	Accuracy	Result
Metal.5d3ubmp0.s1	Metal	1s	0%	1 uncertain	Show classification Show in data acquisition
Metal.5d3u89qk.s16	Metal	1s	100%	1 Metal	Show classification Show in data acquisition
Glass.5d3tpkfb.s12	Glass	1s	100%	1 Glass	Show classification Show in data acquisition
Glass.5d3tqq4c.s11	Glass	1s	100%	1 Glass	Show classification Show in data acquisition
Cardboard.5d3n0e8t	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition
Cardboard.5d3n432g	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition
Cardboard.5d3ndo8j	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition
Cardboard.5d3tkn16.s10	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition
Cardboard.5d3tjdis.s1	Cardboard	1s	0%	1 uncertain	Show classification Show in data acquisition
Metal.5ccr71tl	Metal	1s	100%	1 Metal	Show classification Show in data acquisition
Noise.5d40tk5v.s4	Noise	5s	100%	5 Noise	Show classification Show in data acquisition
Noise.5d40tk5v.s1	Noise	5s	60%	3 Noise, 2 Plastic	Show classification Show in data acquisition
Noise.5d412ku3.s4	Noise	5s	80%	4 Noise, 1 Plastic	Show classification Show in data acquisition
Noise.5d412ku3.s1	Noise	5s	80%	4 Noise, 1 Plastic	Show classification Show in data acquisition
Cardboard.5d3ne92i	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition
Cardboard.5d3n770e	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition
Cardboard.5d3tkn16.s7	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition
Cardboard.5d3tkn16.s13	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition
Cardboard.5d3tjdis.s12	Cardboard	1s	0%	1 Plastic	Show classification Show in data acquisition
Cardboard.5d3tkn16.s18	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition
Cardboard.5d3tkn16.s15	Cardboard	1s	100%	1 Cardboard	Show classification Show in data acquisition

Model testing output

Results

Accuracy

82.64%

Metrics for Classifier

Metric	Value
Area under ROC Curve	0.98
Weighted average Precision	0.88
Weighted average Recall	0.86
Weighted average F1 score	0.86

Confusion matrix

	Cardboard	Glass	Metal	Noise	Plastic	uncertain
Cardboard	70%	0%	5%	0%	10%	15%
Glass	0%	90.9%	3.0%	0%	0%	6.1%
Metal	4.2%	0%	75%	0%	8.3%	12.5%
Noise	0%	0%	0%	80%	20%	0%
Plastic	0%	0%	8.3%	0%	91.7%	0%
f1 score	0.80	0.95	0.78	0.89	0.81

Feature explorer

Error while loading features: {"response":{"readyState":0,"status":0,"statusText":"error"},"errorThrown":""}

These users will be notified over email when this job finishes.

	Name
	MB Mathilde Bindslev