Tracking objects with changing appearances
Implementations are described herein for tracking objects with changing appearances across temporally-disparate images. In various implementations, a first probability distribution over a plurality of classes may be determined for a first biological object depicted in a first image captured at a first point in time. The classes may represent stages of growth of biological objects. Additional probability distribution(s) over the plurality of classes may be determined for candidate biological object(s) depicted in a second image captured at a second point in time subsequent to the first point in time. The candidate biological object(s) may potentially match the first biological object depicted in the first image. Based on a time interval between the first and second points in time, the first probability distribution may be compared to the probability distribution(s) of the candidate biological object(s) depicted in the second image to match one of the candidate biological object(s) depicted in the second image to the first biological object depicted in the first image.
1 . A method implemented using one or more processors, comprising:
determining a first probability distribution over a plurality of classes of a first biological object depicted in a first image captured at a first point in time;
determining one or more additional probability distributions over the plurality of classes of one or more candidate biological objects depicted in a second image, wherein the one or more candidate biological objects depicted in the second image potentially match the first biological object depicted in the first image, and wherein the second image is captured at a second point in time that differs from the first point in time;
predicting a probability distribution over the plurality of classes of the first biological object at the second point in time based at least in part on the probability distribution over the plurality of classes of the first biological object; and
comparing the predicted probability distribution to the one or more additional probability distributions of the one or more candidate biological objects depicted in the second image to match a given biological object of the one or more candidate biological objects depicted in the second image to the first biological object depicted in the first image.
2 . The method of claim 1 , wherein the one or more candidate biological objects include fruits, flowers, or berries, and the plurality of classes include two or more of flowering, ripening, growing, or branching.
3 . The method of claim 1 , wherein the predicting includes processing the probability distribution of the first biological object based on a time interval between the first and second points in time using a probability distribution predictor (PDP) machine learning model, wherein the PDP machine learning model is trained to map temporally disparate probability distributions of biological objects to each other based on time intervals between the temporally disparate probability distributions.
4 . The method of claim 3 , wherein the PDP machine learning model includes a recurrent neural network.
5 . The method of claim 3 , wherein the PDP machine learning model includes a transformer network.
6 . The method of claim 1 , wherein the plurality of classes represent stages of growth of biological objects.
7 . The method of claim 1 , wherein the plurality of classes represent different colors of biological objects.
8 . A system comprising one or more processors and memory storing instructions that, in response to execution by the one or more processors, cause the one or more processors to:
determine a first probability distribution over a plurality of classes of a first biological object depicted in a first image captured at a first point in time;
determine one or more additional probability distributions over the plurality of classes of one or more candidate biological objects depicted in a second image, wherein the one or more candidate biological objects depicted in the second image potentially match the first biological object depicted in the first image, and wherein the second image is captured at a second point in time that differs from the first point in time;
predict a probability distribution over the plurality of classes of the first biological object at the second point in time based at least in part on the probability distribution over the plurality of classes of the first biological object; and
compare the predicted probability distribution to the one or more additional probability distributions of the one or more candidate biological objects depicted in the second image to match a given biological object of the one or more candidate biological objects depicted in the second image to the first biological object depicted in the first image.
9 . The system of claim 8 , wherein the one or more candidate biological objects include fruits, flowers, or berries, and the plurality of classes include two or more of flowering, ripening, growing, or branching.
10 . The system of claim 8 , wherein the instructions to predict include instructions to process the probability distribution of the first biological object based on a time interval between the first and second points in time using a probability distribution predictor (PDP) machine learning model, wherein the PDP machine learning model is trained to map temporally disparate probability distributions of biological objects to each other based on time intervals between the temporally disparate probability distributions.
11 . The system of claim 10 , wherein the PDP machine learning model includes a recurrent neural network.
12 . The system of claim 10 , wherein the PDP machine learning model includes a transformer network.
13 . The system of claim 8 , wherein the plurality of classes represent stages of growth of biological objects.
14 . The system of claim 8 , wherein the plurality of classes represent different colors of biological objects.
15 . At least one non-transitory computer-readable medium comprising instructions that, in response to execution by one or more processors, cause the one or more processors to:
determine a first probability distribution over a plurality of classes of a first biological object depicted in a first image captured at a first point in time;
determine one or more additional probability distributions over the plurality of classes of one or more candidate biological objects depicted in a second image, wherein the one or more candidate biological objects depicted in the second image potentially match the first biological object depicted in the first image, and wherein the second image is captured at a second point in time that differs from the first point in time;
predict a probability distribution over the plurality of classes of the first biological object at the second point in time based at least in part on the probability distribution over the plurality of classes of the first biological object; and
compare the predicted probability distribution to the one or more additional probability distributions of the one or more candidate biological objects depicted in the second image to match a given biological object of the one or more candidate biological objects depicted in the second image to the first biological object depicted in the first image.
16 . The at least one non-transitory computer-readable medium of claim 15 , wherein the one or more candidate biological objects include fruits, flowers, or berries, and the plurality of classes include two or more of flowering, ripening, growing, or branching.
17 . The at least one non-transitory computer-readable medium of claim 15 , wherein the instructions to predict include instructions to process the probability distribution of the first biological object based on a time interval between the first and second points in time using a probability distribution predictor (PDP) machine learning model, wherein the PDP machine learning model is trained to map temporally disparate probability distributions of biological objects to each other based on time intervals between the temporally disparate probability distributions.
18 . The at least one non-transitory computer-readable medium of claim 17 , wherein the PDP machine learning model includes a recurrent neural network.
19 . The at least one non-transitory computer-readable medium of claim 17 , wherein the PDP machine learning model includes a transformer network.
20 . The at least one non-transitory computer-readable medium of claim 15 , wherein the plurality of classes represent stages of growth of biological objects.