Add object detection, tracking, mobile notifications, and search to any security camera.
Settings4 (#173)
Rory Cleard7393b209cff
diff --git a/clearcam.py b/clearcam.pyindex b3d7286..6ee5bfa 100644--- a/clearcam.py+++ b/clearcam.py@@ -276,7 +276,7 @@ class VideoCapture:img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)clip_latest_img(img)process_latest_face(img)- except Exception: print("error in object processing", object_queue[0])+ except Exception as e: print("error in object processing", object_queue[0], e)del object_queue[0]@@ -736,7 +736,7 @@ class HLSRequestHandler(BaseHTTPRequestHandler):returnif parsed_path.path == "/get_global_settings":- self.send_200(global_settings.__dict__)+ self.send_200(secret_settings(global_settings).__dict__)returnif parsed_path.path == "/get_max_storage":self.send_200(body={"max_gb":self.server.max_gb})@@ -1027,6 +1027,9 @@ class HLSRequestHandler(BaseHTTPRequestHandler):content_length = int(self.headers.get('Content-Length', 0))body = self.rfile.read(content_length)data = json.loads(body.decode('utf-8'))+ # keep userid and key if "True"+ if data["userID"] == True: data["userID"] = global_settings.userID+ if data["key"] == True: data["key"] = global_settings.keyadd_to_queue(db.run_put, database, "global_settings", "all", GlobalSettings(**data))add_to_queue(set_settings, GlobalSettings(**data))self.send_200([])@@ -1269,15 +1272,18 @@ def set_settings(x): # todo, save to db, do logic in GlobalSettings class, sanitif x.use_qwen:qwen = Qwen3VL(size=f"{x.qwen_size}B", res=(544, 960)) # h, w. they need to be multiples of 32qwen_prompt = "What has been detected on my CCTV camera? Write in one short sentence"+ print("prewarming Qwen")qwen.prewarm()+ print("DONE")else: qwen = None-global_settings = xdef clip_latest_img(img):if global_settings.use_clip:img = object_finder.preprocess(img)- data = pickle.load(open(object_queue[0].parent / 'embeddings.pkl', 'rb')) if os.path.exists(object_queue[0].parent / 'embeddings.pkl') else {}+ try:+ data = pickle.load(open(object_queue[0].parent / 'embeddings.pkl', 'rb')) if os.path.exists(object_queue[0].parent / 'embeddings.pkl') else {}+ except Exception: data = {}if "embeddings" not in data: data["embeddings"] = {}emb = jit_infer(object_finder.model.precompute_embedding, Tensor(img).unsqueeze(0), jit_cache).numpy()data["embeddings"][str(object_queue[0])] = emb@@ -1390,7 +1396,7 @@ class ThreadedHTTPServer(ThreadingMixIn, HTTPServer):super().server_close()class GlobalSettings:- def __init__(self, use_clip=True, use_face=False ,model_size="t", model_res=960, userID=None, key=None, use_qwen=False, qwen_size=2):+ def __init__(self, use_clip=False, use_face=False ,model_size="t", model_res=960, userID=None, key=None, use_qwen=False, qwen_size=2):self.use_clip = use_clipself.use_face = use_faceself.model_size = model_size@@ -1400,6 +1406,18 @@ class GlobalSettings:self.use_qwen = use_qwenself.qwen_size = qwen_size+def secret_settings(settings):+ return GlobalSettings(+ use_clip=settings.use_clip,+ use_face=settings.use_face,+ model_size=settings.model_size,+ model_res=settings.model_res,+ userID=settings.userID is not None,+ key=settings.key is not None,+ use_qwen=settings.use_qwen,+ qwen_size=settings.qwen_size+ )+if __name__ == "__main__":jit_cache = {}yolo_jit_cache = {}@@ -1409,45 +1427,8 @@ if __name__ == "__main__":cams = database.run_get("links", None)classes = {"0","1","2","7"} # person, bike, car, truck, bird (14)- userID = next((arg.split("=", 1)[1] for arg in sys.argv[1:] if arg.startswith("--userid=")), None)- key = next((arg.split("=", 1)[1] for arg in sys.argv[1:] if arg.startswith("--key=")), None)- use_face = next((arg.split("=", 1)[1] for arg in sys.argv[1:] if arg.startswith("--use_face=")), None)- if use_face: use_face = use_face != "False" # str to bool- model_variant = next((arg.split("=", 1)[1] for arg in sys.argv[1:] if arg.startswith("--yolo_size=")), None)- yolo_res = next((arg.split("=", 1)[1] for arg in sys.argv[1:] if arg.startswith("--yolo_res=")), 640)- if not model_variant:- model_variant = int(input("Select a YOLOv9 model from: \n1: tiny\n2: small\n3: medium\n4: large\n5: xl\nor an RFDETER Model from \n6: nano\n7: small\n8: medium\n9: large\nor press enter to skip (defaults to tiny):") or "1")- if model_variant < 6:- yolo_ress = {"1":320, "2":640, "3":960,"4":1280,"5":1536}- yolo_res = yolo_ress[input("\nSelect a YOLOV9 resoltuion from \n1: 320\n2: 640\n3: 960\n4: 1280\n5: 1536\nor press enter to skip (defaults to 960):") or "3"]- use_clip = input("Would you like to enable clip search on events? (y/n) (1.7GB model), or press enter to skip:") or False- use_clip = use_clip in ["y", "Y"]- use_face = input("Would you like to enable (experimental) face recognition search? (y/n), or press enter to skip:") or False- use_face = use_face in ["y", "Y"]-- userID = input("enter your Clearcam user id or press Enter to skip: ")- use_qwen = False- qwen_size = 2- if len(userID) > 0:- key = ""- while len(key) < 1: key = input("enter a password for encryption: ")- qwen_size = input("Select a Qwen3VL model for AI Summaries from \n2: 2B\n4: 4B\nor press enter to skip:") or False- if qwen_size in ["2", "4"]:- print("prewarming Qwen3VL....")- qwen = Qwen3VL(size=f"{qwen_size}B", res=(544, 960)) # h, w. they need to be multiples of 32- qwen_prompt = "What has been detected on my CCTV camera? Write in one short sentence"- qwen.prewarm()- print("DONE")- use_qwen = True- else: userID = None-- global_settings = GlobalSettings(use_clip=use_clip, model_size=models[model_variant], model_res=yolo_res,- userID=userID, key=key, use_qwen=use_qwen, qwen_size=qwen_size)+ global_settings = GlobalSettings()database.run_put("global_settings", "all", global_settings)-- if userID is not None and key is None:- print("Error: key is required when userID is provided")- sys.exit(1)from models.objects import ObjectFinder@@ -1458,8 +1439,8 @@ if __name__ == "__main__":color_dict = {label: tuple((((i+1) * 50) % 256, ((i+1) * 100) % 256, ((i+1) * 150) % 256)) for i, label in enumerate(class_labels)}cam = None- model = YOLOv9(models[int(model_variant)], res=int(yolo_res)) if int(model_variant) < 6 else RFDETR(models[int(model_variant)])- object_finder = ObjectFinder(clip=global_settings.use_clip, face=use_face)+ model = YOLOv9(global_settings.model_size, res=int(global_settings.model_res))+ object_finder = ObjectFinder()cam = VideoCapture()try:diff --git a/mainview.html b/mainview.htmlindex 7a09b65..89c9e78 100644--- a/mainview.html+++ b/mainview.html@@ -1286,6 +1286,7 @@<body><div id="cameraList" class="camera-grid"></div><div class="multi-view-wrapper">+ <button onclick="openSettingsModal()">Settings</button><button onclick="toggleMultiView()">Multi View</button><div class="date-picker-container"><select id="streamDateSelect" style="padding: 6px; border: 1px solid var(--border-color); border-radius: 4px; font-size: 14px; background: var(--bg-primary); color: var(--text-primary);">@@ -1457,6 +1458,24 @@</div></div></div>++ <div id="settingsModal" class="modal">+ <div class="modal-content" style="max-width: 500px;">+ <div class="modal-header">+ <h3>Global Settings</h3>+ <span class="close" onclick="closeSettingsModal()">×</span>+ </div>++ <div id="settingsContainer" style="display: flex; flex-direction: column; gap: 12px;">+ <p>Loading settings...</p>+ </div>++ <div class="form-actions" style="margin-top: 20px; display: flex; justify-content: center;">+ <button type="button" onclick="saveSettings()">Save</button>+ </div>+ </div>+ </div>+<div id="alertModal" class="modal"><div class="modal-content" style="max-width: 400px;"><div class="modal-header">@@ -3419,6 +3438,189 @@}});+ async function openSettingsModal() {+ const modal = document.getElementById("settingsModal");+ modal.style.display = "flex";++ const container = document.getElementById("settingsContainer");+ container.innerHTML = "<p>Loading settings...</p>";++ try {+ const res = await fetch("/get_global_settings");+ const data = await res.json();++ renderSettings(data);+ } catch (e) {+ container.innerHTML = "<p style='color:red;'>Failed to load settings</p>";+ }+ }++ function closeSettingsModal() {+ document.getElementById("settingsModal").style.display = "none";+ }++ function renderSettings(settings) {+ const container = document.getElementById("settingsContainer");+ container.innerHTML = "";+ Object.entries(settings).forEach(([key, value]) => {++ const labelMap = {+ "use_clip": "OpenCLIP Search",+ "use_face": "Face Detection (Experimental)",+ "model_size": "YOLOv9 Model",+ "model_res": "YOLOv9 Resolution",+ "use_qwen": "AI Event Summaries (Qwen3VL)",+ "qwen_size": "Qwen3VL Size",+ "userID": "User ID",+ "key": "Key"+ };++ const tooltipMap = {+ "use_clip": "Enables searching events using natural language or image similarity with OpenCLIP.",+ "use_face": "Enables searching detected people using a face image.",+ "model_size": "YOLOv9 Detection model size (tiny, small, medium, compact, extended)",+ "model_res": "Sets the input resolution for YOLOv9 detection.",+ "use_qwen": "Enables AI-powered event summaries for notifications using Qwen3VL vision-language model (Clearcam User ID required).",+ "qwen_size": "Select the Qwen3VL model size (2B or 4B parameters).",+ "userID": "Your account user ID for notifications and remote config.",+ "key": "Secret key for encrypting event clips."+ };++ const wrapper = document.createElement("div");+ wrapper.style.display = "flex";+ wrapper.style.justifyContent = "space-between";+ wrapper.style.alignItems = "center";+ wrapper.style.gap = "10px";++ const label = document.createElement("label");+ const labelContent = (labelMap[key] || key);+ label.innerHTML = `+ ${labelContent}+ <span class="tooltip" style="margin-left: 4px;">+ <span class="tooltip-icon">?</span>+ <span class="tooltiptext">${tooltipMap[key] || ''}</span>+ </span>+ `;+ label.style.flex = "1";+ let input;+ if (key === "model_size") {+ input = document.createElement("select");+ ["t", "s", "m", "c", "e"].forEach(opt => {+ const option = document.createElement("option");+ option.value = opt;+ option.textContent = opt;+ if (value === opt) option.selected = true;+ input.appendChild(option);+ });+ }+ else if (key === "model_res") {+ input = document.createElement("select");+ [320, 640, 960, 1280, 1536].forEach(opt => {+ const option = document.createElement("option");+ option.value = opt;+ option.textContent = opt;+ if (value == opt) option.selected = true;+ input.appendChild(option);+ });+ }+ else if (key === "qwen_size") {+ input = document.createElement("select");+ const qwenOptions = [+ { value: 2, label: "2B" },+ { value: 4, label: "4B" }+ ];+ qwenOptions.forEach(opt => {+ const option = document.createElement("option");+ option.value = opt.value;+ option.textContent = opt.label;+ if (value == opt.value) option.selected = true;+ input.appendChild(option);+ });+ }+ else if (key === "userID" || key === "key") {+ input = document.createElement("input");+ input.type = "password";+ input.value = value ? "*****" : "";+ input.placeholder = key;+ input.style.width = "120px";+ }+ else if (typeof value === "boolean") {+ input = document.createElement("input");+ input.type = "checkbox";+ input.checked = value;+ }+ else {+ input = document.createElement("input");+ input.type = "text";+ input.value = value;+ input.style.width = "120px";+ }+ input.dataset.key = key;+ wrapper.appendChild(label);+ wrapper.appendChild(input);+ container.appendChild(wrapper);+ });+ }++ async function saveSettings() {+ const container = document.getElementById("settingsContainer");+ const inputs = container.querySelectorAll("input, select");+ const payload = {};++ inputs.forEach(input => {+ const key = input.dataset.key;+ if (!key) return;++ if (input.type === "checkbox") {+ payload[key] = input.checked;+ }+ else if (input.tagName === "SELECT") {+ const val = input.value;+ if (["model_res", "qwen_size"].includes(key)) {+ payload[key] = Number(val);+ } else {+ payload[key] = val;+ }+ }+ else {+ const val = input.value;+ if (key === "userID" || key === "key") {+ if (val === "*****") {+ payload[key] = true;+ } else if (val === "") {+ payload[key] = null;+ } else {+ payload[key] = val;+ }+ } else {+ const num = Number(val);+ payload[key] = isNaN(num) ? val : num;+ }+ }+ });++ if ((payload.userID && payload.userID !== true) || (payload.key && payload.key !== true)) {+ if (!confirm("Warning: Your User ID and Key will be sent over the network.\n\n" ++ "For your security, avoid entering these on public or unsecured WiFi networks. " ++ "It is recommended to configure this over localhost or a trusted private network.\n\n" ++ "Do you want to continue?")) {+ return;+ }+ }++ const body = JSON.stringify(payload);+ closeSettingsModal();+ fetch("/edit_settings", {+ method: "POST",+ headers: {+ "Content-Type": "application/json"+ },+ body: body+ }).catch(e => {+ console.error("Failed to save settings:", e);+ });+ }+fetchCameras();setInterval(fetchCameras, 5000);</script>diff --git a/models/objects.py b/models/objects.pyindex aa7ecee..f0bc083 100644--- a/models/objects.py+++ b/models/objects.py@@ -186,21 +186,15 @@ def encode_text(model, text):return x / (x * x).sum(axis=-1, keepdim=True).sqrt()class ObjectFinder:- def __init__(self, base_path="data/cameras", clip=False, face=False):+ def __init__(self, base_path="data/cameras"):self.base_path = base_pathself.image_embeddings = {}self.face_embeddings = {}self.image_paths = {}self.face_paths = {}- self.clip = False # todo- self.face = face+ self.clip = False # todo, should be able to init at start+ self.face = False # todoself.jit_cache = {}-- if clip: self.init_clip()-- if self.face:- self.blazeface = BlazeFace()- self.adaface = ADAFACE()def init_clip(self):if self.clip: return