EFEKTIVITAS PENGGUNAAN GOOGLE TEACHABLE MACHINE DALAM PENENTUAN BODY CONDITION SCORE 2 DAN 3 PADA SAPI PASUNDAN
Muhammad Dafa Hidayat, Prof. Ir. Tri Satya Mastuti Widi, S.Pt., M.P., M.Sc., Ph.D., IPM., ASEAN Eng.
2026 | Skripsi | ILMU DAN INDUSTRI PETERNAKAN
Penelitian ini bertujuan untuk menguji efektivitas platform Google Teachable Machine (GTM) dalam mengklasifikasikan Body Condition Score (BCS) 2 dan 3 pada sapi pasundan, serta menganalisis pengaruh berbagai metode pengeditan gambar terhadap akurasi deteksi model. Penelitian menggunakan metode observasional kuantitatif dengan enam variasi perlakuan pengeditan gambar yang meliputi format warna (warna utuh, crop warna, rear warna) dan format skala keabuan (grayscale utuh, crop grayscale, rear grayscale). Parameter yang diamati difokuskan pada nilai akurasi dan simpangan baku (standar deviasi) untuk mengukur tingkat efektivitas dan kestabilan model. Hasil penelitian menunjukkan bahwa jenis pengeditan gambar berpengaruh nyata terhadap performa klasifikasi GTM. Perlakuan grayscale utuh menghasilkan rataan akurasi tertinggi sebesar 77,78% ± 6,36, yang membuktikan bahwa penghapusan warna efektif mengurangi gangguan (noise) visual dan memperjelas morfologi tulang. Sebaliknya, pengambilan gambar dari sudut belakang (rear grayscale) menghasilkan akurasi terendah sebesar 58,33% ± 4,17 akibat terjadi penutupan objek. Hasil penelitian ini menunjukkan bahwa teknologi GTM dapat digunakan sebagai alat bantu objektif dalam sistem pendukung keputusan penilaian BCS sapi pasundan dengan standarisasi pengambilan gambar yang lebih optimal dari sisi samping secara utuh dalam format grayscale.
The purpose of this study is to assess how successfully the Google Teachable Machine (GTM) platform classifies Body Condition Scores (BCS) 2 and 3 in Pasundan cattle and to examine the impact of different image preprocessing techniques. In addition to analyzing the impact of several picture preprocessing techniques on the model's detection accuracy, this study attempts to assess the efficacy of the Google Teachable Machine (GTM) platform in categorizing Body Condition Score (BCS) 2 and 3 in Pasundan cattle. Six picture editing treatments, including color formats (full color, crop color, rear color) and grayscale formats (full grayscale, crop grayscale, rear grayscale), were used in this study's quantitative observational method. The metrics utilized to assess the efficacy were accuracy values and standard deviations.and stability of the model. The findings demonstrated that GTM's classification ability was highly impacted by the kind of image alteration. The full grayscale treatment yielded the highest average accuracy of 77.78% ± 6.36, demonstrating the effective reduction of visual noise and clarification of bone morphological traits with the removal of color elements. On the other hand, because of the object blocking error phenomena, images captured from the rear perspective (rear grayscale) had the lowest accuracy of 58.33% ± 4.17. These findings suggest that GTM technology, with a defined process of obtaining complete side-view photos in grayscale format, can be used as an objective tool in a decision support system for pasundan cattle BCS assessment.
Kata Kunci : Body Condition Score, Machine learning, Google teachable machine, Sapi pasundan, Klasifikasi gambar.