How muhammad-zainal-muttaqin/project-expertise works
This repository is an empirical oil-palm bunch detection and counting research project. Its principal result path is detector proposals, cross-view fusion/linking, bunch-level maturity classification, and per-tree counting, with evaluation of localization and counting quality. It also contains a separate multi-view bunch association and classification subproject. The graph emphasizes these product stages and the newer remote pipeline; evidence is limited to README statements and the supplied source excerpts/index, so unsampled implementations are represented only where the tree or documentation supports them.
Main entry points: eval_endtoend.py End-to-End Evaluation · pipeline_v2.py Four-View Pipeline · penaut_pertandan.py View Linker · train_yolo_4ch_screening.py Detector Training
Components 19
Data and Detection
- RGB+D Preparation
scripts/build_4ch_dataset.py - Detector Training
scripts/train_yolo_4ch_screening.py - Detector Inference
scripts/infer_detektor_damimas.py - Prediction Evaluation
scripts/eval_dump_damimas.py
Remote Pipeline
- Proposal Fusion
scripts/eval_remote_pipeline_postprocess.py - Edge Linker Training
scripts/train_detection_edge_linker.py - Four-View Pipeline
results/remote_eval_2026-08-28/validation_wave/scripts/pipeline_v2.py - Cluster Class Head
results/remote_eval_2026-08-28/validation_wave/scripts/member_head.py - Pipeline Harness
results/remote_eval_2026-08-28/validation_wave/scripts/harness.py - Count Reconciliation
scripts/evaluate_remote_count_reconciled.py - Rank and Emit
results/remote_eval_2026-08-28/scripts/rank_and_emit.py
Bunch Association
- Proposal Classifier
pipeline-pertandan/scripts/classifier_hibrida_damimas.py - View Linker
pipeline-pertandan/scripts/penaut_pertandan.py - Bunch Re-identification
pipeline-pertandan/scripts/reid_pertandan.py - Global Linker
pipeline-pertandan/scripts/linker_global_damimas.py - Per-Tree Counting
pipeline-pertandan/scripts/counting_damimas.py
Evaluation
- End-to-End Evaluation
pipeline-pertandan/scripts/eval_endtoend.py - Locked Remote Evaluation
results/remote_eval_2026-08-28/scripts/run_test_locked.py
Other components
- Researcher
Connections 16
- Researcher → RGB+D Preparation: prepares data
- Researcher → Detector Training: runs experiments
- Detector Training → Detector Inference: trains detectors
- Detector Inference → Prediction Evaluation: produces predictions
- Prediction Evaluation → Proposal Fusion: provides detector dumps
scripts/eval_remote_pipeline_postprocess.py - Edge Linker Training → Four-View Pipeline: trains cross-view linker
results/remote_eval_2026-08-28/validation_wave/scripts/pipeline_v2.py - Four-View Pipeline → Cluster Class Head: classifies clusters
- Four-View Pipeline → Count Reconciliation: reconciles counts
scripts/evaluate_remote_count_reconciled.py - Four-View Pipeline → Pipeline Harness: builds payloads
results/remote_eval_2026-08-28/validation_wave/scripts/harness.py - Locked Remote Evaluation → Rank and Emit: runs evaluation
results/remote_eval_2026-08-28/scripts/run_test_locked.py - Proposal Classifier → End-to-End Evaluation: evaluates classification
pipeline-pertandan/scripts/eval_endtoend.py - View Linker → Bunch Re-identification: supports re-identification
pipeline-pertandan/scripts/reid_pertandan.py - Bunch Re-identification → End-to-End Evaluation: evaluates association
pipeline-pertandan/scripts/eval_endtoend.py - View Linker → End-to-End Evaluation: evaluates linking
pipeline-pertandan/scripts/eval_endtoend.py - Global Linker → End-to-End Evaluation: supports evaluation
pipeline-pertandan/scripts/linker_global_damimas.py - Per-Tree Counting → End-to-End Evaluation: evaluates counting