File size: 4,389 Bytes
15b8951 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 | import numpy as np
from vlm.webconsole.demo1.annotator import (
targets_from_text, draw_dets, LiveAnnotator, ema_box, pick_best,
)
COCO = ["person", "bottle", "chair", "dining table", "cup", "tv", "backpack"]
def test_target_english_bottle():
assert targets_from_text("move to the big blue bottle and stop", COCO) == {"bottle"}
def test_target_vietnamese_chair():
assert targets_from_text("đi tới cái ghế rồi dừng", COCO) == {"chair"}
def test_target_vietnamese_binh_nuoc_to_bottle():
assert targets_from_text("đi tới bình nước", COCO) == {"bottle"}
def test_target_multiple():
got = targets_from_text("find the person sitting on the chair", COCO)
assert got == {"person", "chair"}
def test_target_none_when_no_match():
assert targets_from_text("đi loanh quanh thôi", COCO) == set()
def test_target_ignores_unknown_class():
# 'rocket' không thuộc COCO -> không trả gì.
assert targets_from_text("go to the rocket", COCO) == set()
def test_draw_dets_keeps_shape():
frame = np.zeros((48, 64, 3), dtype=np.uint8)
out = draw_dets(frame, [(5, 5, 30, 30, "bottle", 0.91)])
assert out.shape == frame.shape
def test_draw_dets_empty_ok():
frame = np.zeros((48, 64, 3), dtype=np.uint8)
assert draw_dets(frame, []).shape == frame.shape
class FakeDetector:
names = {0: "person", 1: "bottle", 2: "chair"}
def detect(self, frame, wanted=None):
return [(1, 1, 10, 10, "bottle", 0.9)] if (wanted and "bottle" in wanted) else []
class FakeFrames:
def get_latest_frame(self):
return np.zeros((48, 64, 3), dtype=np.uint8)
# --- lọc nhiễu + EMA -------------------------------------------------------- #
def test_ema_box_first_is_current():
assert ema_box(None, (0, 0, 10, 10), 0.4) == (0, 0, 10, 10)
def test_ema_box_blends():
assert ema_box((0, 0, 0, 0), (10, 10, 10, 10), 0.5) == (5, 5, 5, 5)
def test_pick_best_filters_conf_and_picks_max():
dets = [(0, 0, 5, 5, "bottle", 0.3), (1, 1, 6, 6, "bottle", 0.9),
(2, 2, 7, 7, "bottle", 0.6)]
assert pick_best(dets, 0.45)[5] == 0.9
assert pick_best(dets, 0.95) is None # tất cả dưới ngưỡng -> None
class JumpyDetector:
"""Trả box nhảy lung tung quanh y2≈300 để test EMA làm mượt."""
names = ["bottle"]
def __init__(self, y2_seq):
self.y2 = list(y2_seq)
self.n = 0
def detect(self, frame, wanted=None):
y2 = self.y2[min(self.n, len(self.y2) - 1)]
self.n += 1
return [(100, 100, 200, y2, "bottle", 0.9)]
class Frames720:
def get_latest_frame(self):
return np.zeros((720, 1280, 3), dtype=np.uint8)
def test_update_smooths_jumpy_y2():
ann = LiveAnnotator(Frames720(), JumpyDetector([300, 500, 300, 500]),
smooth_alpha=0.4, min_conf=0.1)
ann.set_target_from_text("go to the bottle")
f = np.zeros((720, 1280, 3), dtype=np.uint8)
ys = []
for _ in range(4):
ann._update_once(f)
ys.append(ann.target_box()[3])
# box thô nhảy 300<->500 nhưng giá trị mượt phải nằm GIỮA, biên độ nhỏ hơn nhiều.
assert all(300 <= y <= 500 for y in ys)
assert max(ys[1:]) - min(ys[1:]) < 200 # dao động bị nén lại
assert ann.frame_height() == 720
def test_update_clears_after_misses():
class NoneDet:
names = ["bottle"]
def detect(self, frame, wanted=None):
return []
ann = LiveAnnotator(Frames720(), NoneDet(), max_miss=2)
ann.set_target_from_text("go to the bottle")
ann.smoothed = (1, 1, 2, 2)
f = np.zeros((720, 1280, 3), dtype=np.uint8)
for _ in range(3):
ann._update_once(f)
assert ann.target_box() is None # mất dấu > max_miss -> xoá box
def test_annotator_render_draws_only_when_target_set():
ann = LiveAnnotator(FakeFrames(), FakeDetector())
frame = np.zeros((48, 64, 3), dtype=np.uint8)
# chưa set target -> trả nguyên frame (không box)
assert np.array_equal(ann.render(frame), frame)
# set target = bottle, nạp box thủ công (bỏ qua thread nền) -> có vẽ
ann.set_target_from_text("đi tới bình nước")
assert ann.targets == {"bottle"}
ann.boxes = ann.detector.detect(frame, wanted=ann.targets)
assert not np.array_equal(ann.render(frame), frame)
|