引導式測驗 · 中等 等級
YOLO即時檢測測驗
回顧 YOLO 的單通道偵測器設計、網格預測、定位權衡和即時幀速率的限制。
問題 1 的 8
在此上下文中,縮寫詞 YOLO 代表什麼?
預覽每個問題8 問題
當您參加測驗時,就會出現答案和解釋。
- 在此上下文中,縮寫詞 YOLO 代表什麼?
- 你只看一次
- 你只學一次
- 您的線上對象定位器
- 產生的物件標籤輸出
- Which cross-domain evaluation did the original YOLO paper report?
- Transcribing speech from noisy recordings
- Generalizing from natural images to artwork datasets
- Classifying satellite images with no labels
- Estimating depth from stereo video only
- 什麼技術可以刪除 YOLO 輸出中重複、重疊的邊界框?
- 批量歸一化
- 輟學
- 非極大值抑制
- 數據增強
- 在最初的YOLO設計中,輸入影像是如何分割進行預測的?
- 進入隨機作物
- 進入顏色通道
- 進入單一區域
- 進入細胞網格
- 每個網格單元與邊界框座標一起預測哪些數量?
- 客觀性置信度得分和類別機率
- 影像解析度
- 相機焦距
- 檔案壓縮比
- 為什麼 YOLO 在視訊和即時攝影應用中特別受歡迎?
- 它無法在 GPU 上運行
- 其單聲道速度可實現即時幀速率
- 這是唯一使用顏色的模型
- 根本不需要任何培訓
- What accuracy tradeoff did the original YOLO paper report relative to leading detection systems?
- It had no localization errors but missed every small object
- It could detect only one class
- It made more localization errors while being less likely to predict false detections where nothing existed
- It required thousands of region-wise classifier runs
- Why can a real-time detector still be unsuitable for a particular deployment?
- A real-time model cannot process images
- Frame rate guarantees safety certification
- A detector’s speed proves every object is found
- Its measured speed does not establish adequate accuracy, latency or robustness for that setting