|
Figure 2. Example images from the IP_RicePests datasets, each image corresponds to a different
species of rice pest: (aBog'liq Classification Method of Significant Rice Pests Ba Физика-курси-1-қисм.-Кодиров-О, ТДИУ, HSyflfAupzrtVxmeX2kSjtbqHjgjAGFqyWgWY1gY, Введение в физику ускорителей заряженных частиц by Иссинский И Б, O\'lchash usullari va vositalari. MamajonovA.A. Sattorov M.O., mustaqil ish 12212121212, Futbol maydonlari va o`lchamlari, TEST argos — копия, «ТЕХНОЛОГИК ЖАРАЁНЛАРНИ АВТОМАТЛАШТИРИШ ВА МОДЕЛЛАШТИРИШ», sinfdan-tashqari-tarbiyaviy-ishlarni-tashkil-qilishning-ilmiy-va-nazariy-asoslariFigure 2. Example images from the IP_RicePests datasets, each image corresponds to a different
species of rice pest: (a) Asiatic rice border; (b) Brown plant hopper; (c) Grain spreader thrips; (d)
Paddy stem maggot; (e) Rice gall midge; (f) Rice leaf caterpillar; (g) Rice leaf hopper; (h) Rice leaf
roller; (i) Rice shell pest; (j) Rice stemfly; (k) Rice water weevil; (l) Small brown plant hopper; (m)
White backed plant hopper; (n) Yellow rice borer.
As can be seen from Figure 2, the IP_RicePests dataset includes 14 types of pests,
including Asiatic rice border, brown plant hopper, grain spreader thrips, paddy stem
maggot, rice gall midge, rice leaf caterpillar, rice leaf hopper, rice leaf roller, rice shell pest,
rice stemfly, rice water weevil, small brown plant hopper, white backed plant hopper and
yellow rice borer. The largest category contained 1110 samples (rice leaf roller) and the
smallest category had only 174 samples (grain spreader thrips). Figure 3 shows the distri-
bution of the number of samples for each pest category in the IP_RicePests dataset, which
shows that the dataset exhibits a natural long-tailed distribution, and clearly demonstrat-
ing that IP_RicePests has a high imbalance rate (IR) [39] in most categories. This is mainly
due to the complexity of the rice field environment and other reasons that make it difficult
to collect samples of individual rice pests, such as grain spreader thrips, paddy stem mag-
got, rice stemfly and other pests, and the number of samples therefore tends to be low.
Unbalanced data can lead to biases in the classification model learning in classes with
relatively more training samples, and therefore, unbalanced data distribution should not
be ignored.
Figure 3. Distribution of the number of samples in each category of the IP_RicePests dataset.
|
|
Bosh sahifa
Aloqalar
Bosh sahifa
Figure 2. Example images from the IP_RicePests datasets, each image corresponds to a different
species of rice pest: (a
|
