Sensor data processing
Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for sensor data processing. The method may include the actions of obtaining sensor data regarding aquatic livestock over periods of time, where the sensor data is captured by at least one sensor at different depths, determining, for each of the periods of time, whether the sensor data captured at different depths during the period of time satisfy one or more evaluation criteria, generating an input data set that concatenates representations of the periods of time, providing the input data set to a machine-learning trained model, receiving, as an output from the machine-learning trained model, an indication of an action to be performed for the aquatic livestock, and initiating performance of the action for the aquatic livestock.
1 . A method comprising:
obtaining sensor data regarding aquatic livestock over periods of time, where the sensor data is captured by at least one sensor while the sensor is positioned at different depths inside an aquatic livestock enclosure;
determining, for each of the periods of time and each of the different depths that the sensor is positioned, whether the sensor data captured while the sensor is positioned at the different depth during the period of time satisfy one or more evaluation criteria;
concatenating matrix representations of the periods of time, each of the matrix representations of the periods of time indicating whether the sensor data captured while the sensor is positioned at the each of the different depths during the period of time satisfy the evaluation criteria;
providing the concatenated matrix representations to a machine-learning trained model;
receiving, as an output from the machine-learning trained model, an indication of an action to be performed for the aquatic livestock based at least on providing the concatenated matrix representations to the machine-learning trained model; and
adjusting an output rate of a feeder device for the aquatic livestock based at least on the indication.
2 . The method of claim 1 , wherein determining, for each of the periods of time, whether the sensor data captured during the period of time satisfy one or more evaluation criteria comprises:
determining whether the sensor data captured at a first depth during a particular period of time satisfies the evaluation criteria; and
determining whether the sensor data captured at a second depth, that is below the first depth, during the particular period of time satisfies the evaluation criteria.
3 . The method of claim 1 , wherein obtaining sensor data regarding aquatic livestock over periods of time, where the sensor data is captured by at least one sensor at different depths inside an aquatic livestock enclosure comprises:
obtaining images of the aquatic livestock captured by a camera at the different depths inside the aquatic livestock enclosure.
4 . The method of claim 1 , wherein obtaining sensor data regarding aquatic livestock over periods of time, where the sensor data is captured by at least one sensor at different depths inside an aquatic livestock enclosure comprises:
obtaining a first portion of the sensor data regarding the aquatic livestock for a first period of time; and
after obtaining the first portion of the sensor data regarding the aquatic livestock for the first period of time, obtaining a second portion of the sensor data regarding the aquatic livestock for a second period of time that is after the first period of time.
5 . The method of claim 1 , wherein the concatenated matrix representations comprises a heat map that represents time along a first dimension and depth along a second dimension, and values at positions in the heat map represent a number of times the evaluation criteria was satisfied at the different depths for each of the periods of time.
6 . The method of claim 5 , wherein columns in the heat map each correspond to a period of time in the periods of time.
7 . The method of claim 1 , wherein receiving, as an output from the machine-learning trained model, an indication of an action to be performed for the aquatic livestock comprises:
receiving an indication to reduce an amount of feed provided to the aquatic livestock.
8 . The method of claim 1 , wherein receiving, as an output from the machine-learning trained model, an indication of an action to be performed for the aquatic livestock comprises:
receiving an indication to increase an amount of feed provided to the aquatic livestock.
9 . The method of claim 1 , comprising:
obtaining training data that includes input data sets that each concatenate representations of previous periods of time and, for each of the input data sets in the training data, a respective indication of a respective action to be performed for the aquatic livestock; and
training the machine-learning trained model with the training data.
10 . The method of claim 1 , wherein the aquatic livestock comprise one or more of fish or crustaceans.
11 . A system comprising:
one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
obtaining sensor data regarding aquatic livestock over periods of time, where the sensor data is captured by at least one sensor while the sensor is positioned at different depths inside an aquatic livestock enclosure;
determining, for each of the periods of time and each of the different depths that the sensor is positioned, whether the sensor data captured while the sensor is positioned at the different depth during the period of time satisfy one or more evaluation criteria;
concatenating matrix representations of the periods of time, each of the matrix representations of the periods of time indicating whether the sensor data captured while the sensor is positioned at the each of the different depths during the period of time satisfy the evaluation criteria;
providing the concatenated matrix representations to a machine-learning trained model;
receiving, as an output from the machine-learning trained model, an indication of an action to be performed for the aquatic livestock based at least on providing the concatenated matrix representations to the machine-learning trained model; and
adjusting an output rate of a feeder device for the aquatic livestock based at least on the indication.
12 . The system of claim 11 , wherein determining, for each of the periods of time, whether the sensor data captured during the period of time satisfy one or more evaluation criteria comprises:
determining whether the sensor data captured at a first depth during a particular period of time satisfies the evaluation criteria; and
determining whether the sensor data captured at a second depth, that is below the first depth, during the particular period of time satisfies the evaluation criteria.
13 . The system of claim 11 , wherein obtaining sensor data regarding aquatic livestock over periods of time, where the sensor data is captured by at least one sensor at different depths inside an aquatic livestock enclosure comprises:
obtaining images of the aquatic livestock captured by a camera at the different depths inside the aquatic livestock enclosure.
14 . The system of claim 11 , wherein obtaining sensor data regarding aquatic livestock over periods of time, where the sensor data is captured by at least one sensor at different depths inside an aquatic livestock enclosure comprises:
obtaining a first portion of the sensor data regarding the aquatic livestock for a first period of time; and
after obtaining the first portion of the sensor data regarding the aquatic livestock for the first period of time, obtaining a second portion of the sensor data regarding the aquatic livestock for a second period of time that is after the first period of time.
15 . The system of claim 11 , wherein the matrix representation comprises a heat map that represents time along a first dimension and depth along a second dimension, and values at positions in the heat map represent a number of times the evaluation criteria was satisfied at the different depths for each of the periods of time.
16 . The system of claim 15 , wherein columns in the heat map each correspond to a period of time in the periods of time.
17 . The system of claim 11 , wherein receiving, as an output from the machine-learning trained model, an indication of an action to be performed for the aquatic livestock comprises:
receiving an indication to reduce an amount of feed provided to the aquatic livestock.
18 . The system of claim 11 , wherein receiving, as an output from the machine-learning trained model, an indication of an action to be performed for the aquatic livestock comprises:
receiving an indication to increase an amount of feed provided to the aquatic livestock.
19 . The system of claim 11 , the operations comprising:
obtaining training data that includes input data sets that each concatenate representations of previous periods of time and, for each of the input data sets in the training data, a respective indication of a respective action to be performed for the aquatic livestock; and
training the machine-learning trained model with the training data.
20 . A computer-readable storage device encoded with a computer program, the program comprising instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:
obtaining sensor data regarding aquatic livestock over periods of time, where the sensor data is captured by at least one sensor while the sensor is positioned at different depths inside an aquatic livestock enclosure;
determining, for each of the periods of time and each of the different depths that the sensor is positioned, whether the sensor data captured while the sensor is positioned at the different depth during the period of time satisfy one or more evaluation criteria;
concatenating matrix representations of the periods of time, each of the matrix representations of the periods of time indicating whether the sensor data captured while the sensor is positioned at the each of the different depths during the period of time satisfy the evaluation criteria;
providing the concatenated matrix representations to a machine-learning trained model;
receiving, as an output from the machine-learning trained model, an indication of an action to be performed for the aquatic livestock based at least on providing the concatenated matrix representations to the machine-learning trained model; and
adjusting an output rate of a feeder device for the aquatic livestock based at least on the indication.