Image Processing Reference
In-Depth Information
Table 10
Recall when Classifying Testing Sets at Three Standard Deviations
Testing as 1Sx (%) 2Sx (%) 3Sx (%)
Fry
67.00
64.50
56.70
Fingerling 86.33
68.35
92.55
Table-fish 92.00
52.00
100.00
Table 11
Precision when Classifying Testing Sets at Three Standard Deviations
Testing as 1Sx (%) 2Sx (%) 3Sx (%)
Fry
73.90
78.84
95.93
Fingerling 63.79
80.00
93.21
Table-fish 54.12
52.00
96.25
FIGURE 7 Recall-precision metrics when classifying fry, fingerling, and table-fish trout at
three standard deviations using our statistical measuring system.
Observing our experimental results when classifying our testing set, we score the best preci-
sion at three standard deviations: 95.93%, 93.21%, and 96.25% for fry, fingerling, and table-fish
trout, respectively.
Furthermore, when classifying those testing lengths, we score a 100% recall and 96.25% pre-
cision when classifying table-fish trout at three standard deviations.
These experimental results are not only motivated, but also valuable evidences that indicate
efectiveness in our classification system.
6 Conclusions
In this article, we have robustly evaluated our statistical system to measure rainbow trout in
farm using computer vision as observed in Figure 8 . This novel technique [ 8 ] is a simple but
efective statistical method, which has been evaluated in a small farm in central Mexico named
Rincon del Sol [ 9 ].
 
 
 
Search WWH ::




Custom Search