00000ctmaa22000001a 4500 UP-1685675941123958486 Buklod 20190724153757.0 a r |||| u| ta 190724s xx d r |||| u| (iLib)UPTAC-00019221841 UPVTC eng LG 993.5 2019 C66 H59 Hizon, Daniel Jr. L. Pollen viability analysis using extremely randomized trees on selected rice (Oryza sativa L.) genotypes. Tacloban City Daniel L. Hizon Jr. 2019 xi, 85 leaves illustrations. Undergraduate thesis (B.S. Computer Science) -- University of the Philippines, Tacloban. Object counting is a common task specially in medicine and laboratories. Tasks such as cell counting is done by manually counting these small species either directly or plotting marks on images taken under the microscope. Pollen viability detection and counting is not a different problem. Manual counting generally, is known to be time consuming and a labor intensive job. A series of validation of output data is needed to achieve higher result accuracy. This limits the research progress in one way or another as it takes considerable amounts of time and resources. Furthermore, certain level of knowledge and expertise is needed for the result not to be compromised. This study used sliding Window approach along with Extremely Randomized Trees (ERT) to solve this problem. To test the accuracy of Sliding window and ERT, two sets of stained pollen grain images were taken under the microscope with both set having 50 images each. The first set are the images without lumped pollen grains and the second set are images with lumped pollen grains. Training set for building the ensemble was extracted form the data set through interactive labeling. Results have verified the feasibility of using ERT in pollen viability classification specially on images without lumped pollen grains. Experiments over different values of parameters were also conducted over images without lumped pollen grains. The results have shown to have a method that can produce an outcome comparable to manual counting. This study has presented a method that may provide an alternative to a time consuming and labor intensive manual counting. FI UP UPTAC UPTAC LG 993.5 2019 C66 H59 Thesis