urn:nasa:pds:mslnet:data:mslnet_predicted_classes
1.0
MSLNet Predicted Class Set
1.17.0.0
Product_Observational
Kiri Wagstaff, Steven Lu, Kevin Grimes, Jesse Cai,
Raymond Francis, and Jake Lee.
2020
MSL Class Predictions generated from analysis of
the MSL surface data set, version 2. The five columns in this
file contain image file names, predicted class in integer
representation, predicted class in string representation,
predicted class probability, and version of the classifier that
made the prediction.
2022-01-07
1.0
Initial version.
The MSL Class Predictions were generated from analysis of
MSL surface data, specifically, surface images from Mastcam and
MAHLI.
2012-08-07Z
2019-07-27Z
Science
Derived
MSL image class predictions capture the semantic
content of images and can be used to enable content-based search
for relevant images.
Mars Science Laboratory
Mission
urn:nasa:pds:context:investigation:mission.mars_science_laboratory
data_to_investigation
Mast Camera
Instrument
urn:nasa:pds:context:instrument:mast_left.msl
is_instrument
Mast Camera
Instrument
urn:nasa:pds:context:instrument:mast_right.msl
is_instrument
Mars Hand Lens Imager Camera
Instrument
urn:nasa:pds:context:instrument:mahli.msl
is_instrument
Curiosity Rover
Host
urn:nasa:pds:context:instrument_host:spacecraft.msl
is_instrument_host
Mars
Planet
urn:nasa:pds:context:target:planet.mars
data_to_target
2.0
MSLNet
Supervised_Learning
Convolutional_Neural_Network_Classifier
Fine-tuned AlexNet (ImageNet) with
Temperature Scaling Calibration
10.5555/2999134
Krizhevsky, A., Sutsekver, I., and
Hinton, G.E., "Imagenet classification with deep
convolutional neural networks," Advances in Neural
Information Processing Systems 25, p., 1097-1105, 2012.
AlexNet citation
1.0
2000
1.0
300
1.0
600
Test accuracy
0.745
Test accuracy with abstention below
0.9 confidence
0.903
Test abstention rate with 0.9
confidence threshold
0.518
mslnet_predicted_classes.csv
2022-01-07T15:10:00.000Z
8719123
136967
MSL Surface Data Class Predictions
TABLE
0
PDS DSV 1
136966
Carriage-Return Line-Feed
Comma
5
0
73
image_filename
1
ASCII_File_Name
34
This field contains image file names from the
Mars Surface Data Set.
predicted_class_id
2
ASCII_NonNegative_Integer
3
This field contains the associated predicted
class of an image file in integer form.
predicted_class_string
3
ASCII_String
21
This field contains the associated predicted
class of an image file in string representation.
predicted_probability
4
ASCII_Real
6
This field contains the predicted probability
for classes.
classifier_version
5
ASCII_String
3
This field contains the version number of the
classifier used to predict the class.