4 Channel Tracking
4 Channel Tracking
Fast In-Plane Out-of-plane Out-of-view Background Low Illumination Scale Deformation Motion Blur
Motion Rotation Rotation Clutters Resolution Variation Variation Occlusion
RGBD Image
Depth Image
Correlation Filter /
Output Cosine Similarity Calculation
Checklist Overview
Literature Review ✓ Literature Overview
✓ Initial Experiment
Output ❏ Visualisasi
❏ Evaluasi
❏ Pembentukan kesimpulan
Literature Review
✓ Literature Overview
Detection result (1 image Detection result (sequence 189
✓ Initial Experiment random) image)
Metodologi Penelitian
Output Tracker Initialization HOG Feature Extraction
Depth Input RGB
[1024, 1024, 4]
RGBD 4-Channel Next Frame Output
Tracking Method:
Cosine Similarity /
Correlation Filter
Progress Milestone
Input: 4-Channel Detection Method: Output
RGBD Image 4-Channel Mask RCNN
Tracking Method:
Cosine Similarity /
Correlation Filter
Input: 4-Channel Persiapan Dataset
RGBD Image
❏ Persiapan dataset RGB dan Depth
❏ Pembentukan groundtruth dan annotasi
gambar
❏ Pembentukan image 4-Channel RGBD dengan
melakukan merge image RGB dan Depth
❏ Visualisasi hasil masking
Persiapan Dataset
Tracking Method:
Cosine Similarity /
Correlation Filter
Detection Method: Pembentukan Learning Model
4-Channel Mask
RCNN ❏ Pembentukan model
❏ Training Model
❏ Hasil Evaluasi
Pembentukan Learning Model
❏ Training Model
❏ Hasil Evaluasi
Pembentukan Learning Model
✓ Ekstraksi Fitur ●
●
Val_Loss: 0.1726
Loss: 0.1218
✓ Training Model ● Preview:
● mAP: 0.9740112999738273
Progress Milestone
Input: 4-Channel Detection Method: Output
RGBD Image 4-Channel Mask RCNN
Tracking Method:
Cosine Similarity
Tracking Method: Aplikasi Algoritma Tracking
Cosine Similarity /
Correlation Filter ❏ Ekstraksi Fitur
❏ Perhitungan Cosine Similarity
❏ Hasil Evaluasi
Aplikasi Algoritma Tracking
❏ Perhitungan Cosine
Similarity
❏ Hasil Evaluasi