Pencemaran sampah plastik di Sungai Sekanak Kota Palembang merupakan salah satu permasalahan lingkungan yang memerlukan pemantauan secara efektif untuk mendukung upaya pelestarian lingkungan. Perkembangan teknologi computer vision memungkinkan proses deteksi objek dilakukan secara otomatis, sehingga dapat dimanfaatkan untuk mengidentifikasi sampah plastik di perairan. Penelitian ini bertujuan membangun model deteksi sampah plastik menggunakan framework YOLOv11 serta menganalisis pengaruh variasi resolusi citra terhadap kemampuan deteksi objek. Dataset penelitian diperoleh dari hasil ekstraksi video menggunakan kamera iPhone 15 yang menghasilkan 652 citra dengan empat kelas objek, yaitu botol plastik, cup plastik, kantong plastik, dan styrofoam. Tahapan penelitian meliputi pre-processing, anotasi dataset, augmentasi data, pembagian dataset, pelatihan model, dan pengujian menggunakan variasi resolusi 240p, 360p, 480p, 720p, dan 1080p. Model terbaik diperoleh pada epoch ke-250 dan digunakan sebagai model akhir (best.pt) dalam seluruh proses pengujian. Hasil penelitian menunjukkan bahwa variasi resolusi citra memengaruhi kemampuan deteksi dan efisiensi komputasi model. Resolusi 1080p menghasilkan akurasi deteksi tertinggi sebesar 89,19%, sedangkan resolusi 360p memberikan waktu inferensi tercepat sebesar 0,7 ms. Selain resolusi citra, kondisi lingkungan pengujian juga memengaruhi performa deteksi objek. Berdasarkan hasil penelitian, framework YOLOv11n mampu mendeteksi sampah plastik dengan baik, dan resolusi 1080p direkomendasikan apabila mengutamakan akurasi, sedangkan resolusi 360p dapat dipilih apabila efisiensi waktu komputasi menjadi prioritas.
Kata Kunci: YOLOv11n, deteksi sampah plastik, resolusi citra, computer vision, Sungai Sekanak.

Plastic waste pollution in the Sekanak River, Palembang City, has become an environmental issue that requires effective monitoring to support environmental conservation efforts. Advances in computer vision technology enable automatic object detection, making it a promising approach for identifying plastic waste in river environments. This study aims to develop a plastic waste detection model using the YOLOv11 framework and to analyze the effect of image resolution on object detection performance. The dataset was obtained from video recordings captured using an iPhone 15 camera and consisted of 652 images representing four object classes: plastic bottles, plastic cups, plastic bags, and styrofoam. The research stages included pre-processing, dataset annotation, data augmentation, dataset splitting, model training, and performance evaluation using five image resolutions: 240p, 360p, 480p, 720p, and 1080p. The best-performing model was achieved at epoch 250 and was selected as the final model (best.pt) for all testing scenarios. The experimental results demonstrate that image resolution significantly influences both detection accuracy and computational efficiency. The 1080p resolution achieved the highest detection accuracy of 89.19%, while the 360p resolution produced the fastest inference time of 0.7 ms. In addition to image resolution, environmental conditions also affected the model's detection performance. Overall, the YOLOv11n framework successfully detected plastic waste in the Sekanak River. The 1080p resolution is recommended for applications requiring higher detection accuracy, whereas the 360p resolution is a suitable alternative when computational efficiency is prioritized.
Keywords: YOLOv11n, plastic waste detection, image resolution, computer vision, Sekanak River.