Plant treatment model selection based on agricultural image interaction
Embodiments relate to selecting and utilizing a plant treatment model. A control system may provide images of plants of different dimensions and types for display to a user. The control system may generate one or more plant treatment action preferences of the user based on user interactions with the provided images of plants. The control system may apply trained plant treatment models to the images of plants, each plant treatment model identifying one or more plants in the images and one or more corresponding plant treatment actions. The control system may select a plant treatment model based on a comparison of (1) plants and corresponding plant treatment actions identified by the plant treatment models in the images and (2) the one or more plant treatment action preferences of the user. The control system may configure a farming machine to operate based on the selected plant treatment model.
1 . A method for selecting a plant treatment model to be used by a farming machine configured to perform treatment actions on plants growing in a field, the method comprising:
providing images of plants of different dimensions and types for display to a user;
generating one or more plant treatment action preferences of the user for the farming machine based on interactions by the user with the provided images of plants;
applying a plurality of trained plant treatment models to the images of plants, each trained plant treatment model identifying one or more plants in the images and one or more corresponding plant treatment actions to be applied to the identified one or more plants;
selecting the plant treatment model from the plurality of trained plant treatment models to be used by the farming machine to perform treatment actions on plants in the field, the plant treatment model selected based on a comparison of (1) plants and corresponding plant treatment actions identified by the plurality of trained plant treatment models in the images and (2) the one or more plant treatment action preferences of the user; and
configuring the farming machine to perform treatment actions on plants in the field using the plant treatment model selected from the plurality of trained plant treatment models.
2 . The method of claim 1 , further comprising: adjusting parameters of the selected plant treatment model based on the one or more plant treatment action preferences of the user.
3 . The method of claim 1 , further comprising:
while the farming machine is operating in the field;
providing indications of plant treatment actions performed by the farming machine to the user;
generating one or more updated plant treatment action preferences based on the interactions by the user with the indications; and
adjusting parameters of the selected plant treatment model based on the updated plant treatment action preferences.
4 . The method of claim 1 , wherein the plurality of trained plant treatment models includes trained plant treatment models trained to identify plants of specific types or dimensions.
5 . The method of claim 1 , wherein the plurality of trained plant treatment models includes trained plant treatment models with different plant identification sensitivity levels.
6 . The method of claim 1 , wherein selecting the plant treatment model from the plurality of trained plant treatment models is further based on a current growth stage of plant crops growing in the field.
7 . The method of claim 1 , wherein the images include crop plants in a same growth stage and weed plants in different growth stages.
8 . The method of claim 1 , wherein the comparison includes determining differences between (1) plants and corresponding plant treatment actions identified by the plurality of trained plant treatment models and (2) the one or more plant treatment action preferences of the user.
9 . The method of claim 8 , wherein differences between (1) the plants and corresponding plant treatment actions identified by the selected plant treatment model and (2) the one or more plant treatment action preferences of the user are at least one of: (a) less than a threshold difference or (b) smaller than the differences between (1) plants and corresponding plant treatment actions identified by the other plant treatment models and (2) the one or more plant treatment action preferences of the user.
10 . The method of claim 1 , wherein the interactions by the user with the provided images of plants include the user manually annotating one or more of the provided images.
11 . A non-transitory computer-readable storage medium comprising stored instructions that, when executed by a computing device, cause the computing device to perform operations including:
providing images of plants of different dimensions and types for display to a user;
generating one or more plant treatment action preferences of the user for a farming machine based on interactions by the user with the provided images of plants;
applying a plurality of trained plant treatment models to the images of plants, each trained plant treatment model identifying one or more plants in the images and one or more corresponding plant treatment actions to be applied to the identified one or more plants;
selecting a plant treatment model from the plurality of trained plant treatment models to be used by the farming machine to perform treatment actions on plants in a field, the plant treatment model selected based on a comparison of (1) plants and corresponding plant treatment actions identified by the plurality of trained plant treatment models in the images and (2) the one or more plant treatment action preferences of the user; and
configuring the farming machine to perform treatment actions on plants in the field using the selected plant treatment model selected from the plurality of trained plant treatment models.
12 . The non-transitory computer-readable storage medium of claim 11 , further comprising: adjusting parameters of the selected plant treatment model based on the one or more plant treatment action preferences of the user.
13 . The non-transitory computer-readable storage medium of claim 11 , further comprising:
while the farming machine is operating in the field;
providing indications of plant treatment actions performed by the farming machine to the user;
generating one or more updated plant treatment action preferences based on the interactions by the user with the indications; and
adjusting parameters of the selected plant treatment model based on the updated plant treatment action preferences.
14 . The non-transitory computer-readable storage medium of claim 11 , wherein the plurality of trained plant treatment models includes trained plant treatment models trained to identify plants of specific types or dimensions.
15 . The non-transitory computer-readable storage medium of claim 11 , wherein the plurality of trained plant treatment models includes trained plant treatment models with different plant identification sensitivity levels.
16 . The non-transitory computer-readable storage medium of claim 11 , wherein selecting the plant treatment model from the plurality of trained plant treatment models is further based on a current growth stage of plant crops growing in the field.
17 . The non-transitory computer-readable storage medium of claim 11 , wherein the images include crop plants in a same growth stage and weed plants in different growth stages.
18 . The non-transitory computer-readable storage medium of claim 11 , wherein the comparison includes determining differences between (1) plants and corresponding plant treatment actions identified by the plurality of trained plant treatment models and (2) the one or more plant treatment action preferences of the user.
19 . The non-transitory computer-readable storage medium of claim 18 , wherein differences between (1) the plants and corresponding plant treatment actions identified by the selected plant treatment model and (2) the one or more plant treatment action preferences of the user are at least one of: (a) less than a threshold difference or (b) smaller than the differences between (1) plants and corresponding plant treatment actions identified by the other plant treatment models and (2) the one or more plant treatment action preferences of the user.
20 . The non-transitory computer-readable storage medium of claim 11 , wherein the interactions by the user with the provided images of plants include the user manually annotating one or more of the provided images.
21 . The method of claim 1 , wherein the images of plants are not images of plants in the field.
22 . The method of claim 1 , wherein the providing the images of plants, the generating one or more plant treatment action preferences of the user, the applying the plurality of trained plant treatment models to the images of plants, and the selecting the plant treatment model from the plurality of trained plant treatment models are performed prior to the farming machine performing treatment actions on plants growing in the field.
23 . The method of claim 1 , further comprising:
subsequent to configuring the farming machine to perform treatment actions on plants in the field using the plant treatment model selected from the plurality of trained plant treatment models, performing, by the farming machine, treatment actions on plants growing in the field using the plant treatment model selected from the plurality of trained plant treatment models.