Underwater camera biomass prediction aggregation
Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for underwater camera biomass prediction aggregation. In some implementations, an exemplary method includes obtaining images of fish captured by an underwater camera; providing data of the images to a trained model; obtaining output of the trained model indicating the likelihoods that the biomass of fish are within multiple ranges; combining likelihoods of the output based on one or more ranges common to likelihoods of two or more fish to generate a biomass distribution; and determining an action based on the biomass distribution.
1. A computer-implemented method comprising:
obtaining images of multiple fish that were generated by an underwater camera;
for each of two or more of the multiple fish, generating a set of features corresponding to the fish using the obtained images;
for each of the two or more of the multiple fish, providing the set of features corresponding to the fish to a model that is trained, based on a given input set of features for a given fish, to output a distribution indicating likelihoods of a biomass of the given fish being within multiple biomass ranges;
for each of the two or more of the multiple fish, receiving a distribution indicating the likelihoods of the biomass of the fish being with the multiple biomass ranges;
aggregating the distributions of the two or more of the multiple fish; and
performing an action based on the aggregated distributions of the two or more of the multiple fish.
2. The method of claim 1 , wherein performing the action comprises transmitting a control message to an actuator, blower, or conveyer of a feed system.
3. The method of claim 1 , wherein the set of features comprises data indicating lengths between key points of each fish.
4. The method of claim 1 , wherein aggregating the distributions comprises summing the likelihoods for each of the multiple biomass ranges.
5. The method of claim 1 , wherein performing the action comprises identifying subpopulations of fish within the aggregated distributions.
6. The method of claim 1 , wherein performing the action comprises changing a location of an ongoing fish feeding session.
7. The method of claim 1 , wherein generating the set of features comprise performing a 3D pose estimation for each of the multiple fish.
8. A non-transitory, computer-readable medium storing one or more instructions executable by a computer system to perform operations comprising:
obtaining images of multiple fish that were generated by an underwater camera;
for each of two or more of the multiple fish, generating a set of features corresponding to the fish using the obtained images;
for each of the two or more of the multiple fish, providing the set of features corresponding to the fish to a model that is trained, based on a given input set of features for a given fish, to output a distribution indicating likelihoods of a biomass of the given fish being within multiple biomass ranges;
for each of the two or more of the multiple fish, receiving a distribution indicating the likelihoods of the biomass of the fish being with the multiple biomass ranges;
aggregating the distributions of the two or more of the multiple fish; and
performing an action based on the aggregated distributions of the two or more of the multiple fish.
9. The medium of claim 8 , wherein performing the action comprises transmitting a control message to an actuator, blower, or conveyer of a feed system.
10. The medium of claim 8 , wherein the set of features comprises data indicating lengths between key points of each fish.
11. The medium of claim 8 , wherein aggregating the distributions comprises summing the likelihoods for each of the multiple biomass ranges.
12. The medium of claim 8 , wherein performing the action comprises identifying subpopulations of fish within the aggregated distributions.
13. The medium of claim 8 , wherein performing the action comprises changing a location of an ongoing fish feeding session.
14. The medium of claim 8 , wherein generating the set of features comprise performing a 3D pose estimation for each of the multiple fish.
15. A computer-implemented system, comprising:
one or more computers; and
one or more computer memory devices interoperably coupled with the one or more computers and having tangible, non-transitory, machine-readable media storing one or more instructions that, when executed by the one or more computers, perform one or more operations comprising:
obtaining images of multiple fish that were generated by an underwater camera;
for each of two or more of the multiple fish, generating a set of features corresponding to the fish using the obtained images;
for each of the two or more of the multiple fish, providing the set of features corresponding to the fish to a model that is trained, based on a given input set of features for a given fish, to output a distribution indicating likelihoods of a biomass of the given fish being within multiple biomass ranges;
for each of the two or more of the multiple fish, receiving a distribution indicating the likelihoods of the biomass of the fish being with the multiple biomass ranges;
aggregating the distributions of the two or more of the multiple fish; and
performing an action based on the aggregated distributions of the two or more of the multiple fish.
16. The system of claim 15 , wherein performing the action comprises transmitting a control message to an actuator, blower, or conveyer of a feed system.
17. The system of claim 15 , wherein the set of features comprises data indicating lengths between key points of each fish.
18. The system of claim 15 , wherein aggregating the distributions comprises summing the likelihoods for each of the multiple biomass ranges.
19. The system of claim 15 , wherein performing the action comprises identifying subpopulations of fish within the aggregated distributions.
20. The system of claim 15 , wherein performing the action comprises changing a location of an ongoing fish feeding session.