Supervisor

supervisor.ex

defmodule MyApp.Worker.Supervisor do
  use Supervisor

  def start_link do
    Supervisor.start_link(__MODULE__, [], name: __MODULE__)
  end

  def init(_) do
    Task.start_link(&MyApp.Worker.Supervisor.initiate_workers/0)

    supervise(
      [worker(MyApp.Worker, [], [restart: :permanent, shutdown: :infinity])],
      strategy: :simple_one_for_one    )
  end

  def start_child(args) do
    Supervisor.start_child(__MODULE__, [args])
  end

  def initiate_workers do
    Camera
    |> MyApp.Repo.all
    |> Enum.map(&(start_camera_worker &1))
  end

  def start_camera_worker(camera_with_index) do
    ....
    .....
    MyApp.Worker.Supervisor.start_child(args)
  end
end

worker.ex

defmodule MyApp.Worker do
  import MyApp.Snapshot
  import MyApp.Schedule

  def start_link(args) do
    worker_name = args[:camera_exid] |> String.to_atom
    GenServer.start_link(__MODULE__, args, name: worker_name)
  end

  def init(args) do
    Task.start_link(fn -> loop(args) end)
  end

  defp loop(args) do
    ...
    ...
  end
end

The approach would be to create a module that would be your WorkerMonitor (feel free to rename it to something more fitting). This WorkerMonitor will be directly supervised by your Supervisor. In your WorkerMonitor genserver, you start the Genserver module that you will be monitoring. I think you could do it with http://elixir-lang.org/docs/master/elixir/Kernel.html#spawn_monitor/3.

In your WorkerMonitor, define a handle_info callback that handles the message {:DOWN, ref, process, pid, reason} then either spawn_monitor/3 the process or do nothing depending on the type of error you got.